title
stringlengths
10
147
abstract
stringlengths
3
2.41k
tldr_text
stringlengths
96
508
content_markdown
stringlengths
0
464k
authors
listlengths
0
41
date
timestamp[ms]
publish_info
stringclasses
115 values
publish_is_top
bool
2 classes
citation_count
uint32
0
1k
citation_count_filtered_math_and_top_conf
uint32
0
130
theorem_provers
listlengths
1
4
url
stringlengths
31
152
arxiv_url
stringlengths
32
32
semantics_scholar_url
stringlengths
78
78
Large Language Models are Zero-Shot Reasoners
Pretrained large language models (LLMs) are widely used in many sub-fields of natural language processing (NLP) and generally known as excellent few-shot learners with task-specific exemplars. Notably, chain of thought (CoT) prompting, a recent technique for eliciting complex multi-step reasoning through step-by-step a...
Experimental results demonstrate that the Zero-shot-CoT, using the same single prompt template, significantly outperforms zero-shot LLM performances on diverse benchmark reasoning tasks including arithmetics, symbolic reasoning, and other logical reasoning tasks, without any hand-crafted few-shot examples.
## Large Language Models are Zero-Shot Reasoners **Takeshi Kojima** The University of Tokyo ``` t.kojima@weblab.t.u-tokyo.ac.jp ``` **Shixiang Shane Gu** Google Research, Brain Team **Machel Reid** **Yutaka Matsuo** **Yusuke Iwasawa** Google Research[∗] The University of Tokyo The University of Tokyo **Abstract*...
[ "Yutaka, Matsuo", "Takeshi, Kojima", "Shixiang Shane, Gu", "Machel, Reid", "Yusuke, Iwasawa" ]
2022-05-24T00:00:00
NeurIPS 2022 Poster
true
1,000
75
null
https://arxiv.org/abs/2205.11916v4
https://arxiv.org/abs/2205.11916
https://www.semanticscholar.org/paper/e7ad08848d5d7c5c47673ffe0da06af443643bda
Measuring Massive Multitask Language Understanding
We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent mode...
While most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average, however, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy.
## MEASURING MASSIVE MULTITASK LANGUAGE UNDERSTANDING **Dan Hendrycks** **Collin Burns** **Steven Basart** **Andy Zou** UC Berkeley Columbia University UChicago UC Berkeley **Mantas Mazeika** **Dawn Song** **Jacob Steinhardt** UIUC UC Berkeley UC Berkeley ABSTRACT We propose a new test to measure a text model’s m...
[ "Dan, Hendrycks", "Andy, Zou", "Collin, Burns", "Dawn, Song", "Mantas, Mazeika", "Steven, Basart", "Jacob, Steinhardt" ]
2021-01-12T00:00:00
ICLR 2021
true
1,000
23
null
http://arxiv.org/abs/2009.03300
https://arxiv.org/abs/2009.03300
https://www.semanticscholar.org/paper/814a4f680b9ba6baba23b93499f4b48af1a27678
ReAct: Synergizing Reasoning and Acting in Language Models
While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we ...
The use of LLMs are explored to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources ...
## REACT: SYNERGIZING REASONING AND ACTING IN LANGUAGE MODELS Shunyu Yao*,1[∗], Jeffrey Zhao2, Dian Yu2, Nan Du2, Izhak Shafran2, Karthik Narasimhan1, Yuan Cao2 1Department of Computer Science, Princeton University 2Google Research, Brain team 1{shunyuy,karthikn}@princeton.edu 2{jeffreyzhao,dianyu,dunan,izhak,yuanca...
[ "Shunyu, Yao", "Karthik, Narasimhan", "Dian, Yu", "Jeffrey, Zhao", "Izhak, Shafran", "Yuan, Cao", "Nan, Du" ]
2023-03-09T00:00:00
ICLR 2023
true
1,000
31
null
http://arxiv.org/abs/2210.03629
https://arxiv.org/abs/2210.03629
https://www.semanticscholar.org/paper/99832586d55f540f603637e458a292406a0ed75d
Self-Consistency Improves Chain of Thought Reasoning in Language Models
Chain-of-thought prompting combined with pretrained large language models has achieved encouraging results on complex reasoning tasks. In this paper, we propose a new decoding strategy, self-consistency, to replace the naive greedy decoding used in chain-of-thought prompting. It first samples a diverse set of reasoning...
This paper proposes a new decoding strategy, self-consistency, to replace the naive greedy decoding used in chain-of-thought prompting that first samples a diverse set of reasoning paths instead of only taking the greedy one, and then selects the most consistent answer by marginalizing out the sampled reasoning paths.
## SELF-CONSISTENCY IMPROVES CHAIN OF THOUGHT REASONING IN LANGUAGE MODELS **Xuezhi Wang[†‡]** **Jason Wei[†]** **Dale Schuurmans[†]** **Quoc Le[†]** **Ed H. Chi[†]** **Sharan Narang[†]** **Aakanksha Chowdhery[†]** **Denny Zhou[†§]** _†Google Research, Brain Team_ _‡xuezhiw@google.com, §dennyzhou@google.com_ ABSTR...
[ "Jason, Wei", "Aakanksha, Chowdhery", "Xuezhi, Wang", "Denny, Zhou", "Dale, Schuurmans", "Ed, Chi", "Quoc, Le", "Sharan, Narang" ]
2023-03-07T00:00:00
ICLR 2023
true
1,000
90
null
http://arxiv.org/abs/2203.11171
https://arxiv.org/abs/2203.11171
https://www.semanticscholar.org/paper/5f19ae1135a9500940978104ec15a5b8751bc7d2
Toolformer: Language Models Can Teach Themselves to Use Tools
Language models (LMs) exhibit remarkable abilities to solve new tasks from just a few examples or textual instructions, especially at scale. They also, paradoxically, struggle with basic functionality, such as arithmetic or factual lookup, where much simpler and smaller specialized models excel. In this paper, we show ...
This paper introduces Toolformer, a model trained to decide which APIs to call, when to call them, what arguments to pass, and how to best incorporate the results into future token prediction, which achieves substantially improved zero-shot performance across a variety of downstream tasks.
## Toolformer: Language Models Can Teach Themselves to Use Tools **Timo Schick** **Jane Dwivedi-Yu** **Roberto Dessì[†]** **Roberta Raileanu** **Maria Lomeli** **Luke Zettlemoyer** **Nicola Cancedda** **Thomas Scialom** Meta AI Research _[†]Universitat Pompeu Fabra_ **Abstract** Language models (LMs) exhibit rem...
[ "Timo, Schick", "Jane, Dwivedi-Yu", "Roberto, Dessì", "Roberta, Raileanu", "Maria, Lomeli", "Luke, Zettlemoyer", "Nicola, Cancedda", "Thomas, Scialom" ]
2023-02-09T00:00:00
NeurIPS 2023 Oral
true
1,000
23
null
http://arxiv.org/abs/2302.04761
https://arxiv.org/abs/2302.04761
https://www.semanticscholar.org/paper/53d128ea815bcc0526856eb5a9c42cc977cb36a7
Training Verifiers to Solve Math Word Problems
State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word pro...
It is demonstrated that verification significantly improves performance on GSM8K, and there is strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline.
## Training Verifiers to Solve Math Word Problems **Karl Cobbe[∗]** **Vineet Kosaraju[∗]** **Mohammad Bavarian** **Mark Chen** **Heewoo Jun** **Lukasz Kaiser** **Matthias Plappert** **Jerry Tworek** **Jacob Hilton** **Reiichiro Nakano** **Christopher Hesse** **John Schulman** OpenAI **Abstract** State-of-the-ar...
[ "Mark, Chen", "John, Schulman", "Reiichiro, Nakano", "Jerry, Tworek", "Vineet, Kosaraju", "Jacob, Hilton", "Heewoo, Jun", "Karl, Cobbe", "Lukasz, Kaiser", "Mohammad, Bavarian", "Matthias, Plappert", "Christopher, Hesse" ]
2021-11-17T00:00:00
null
false
1,000
130
null
http://arxiv.org/abs/2110.14168
https://arxiv.org/abs/2110.14168
https://www.semanticscholar.org/paper/d6045d2ccc9c09ca1671348de86d07da6bc28eea
Tree of Thoughts: Deliberate Problem Solving with Large Language Models
Language models are increasingly being deployed for general problem solving across a wide range of tasks, but are still confined to token-level, left-to-right decision-making processes during inference. This means they can fall short in tasks that require exploration, strategic lookahead, or where initial decisions pla...
A new framework for language model inference, Tree of Thoughts (ToT), which generalizes over the popular Chain of Thought approach to prompting language models, and enables exploration over coherent units of text (thoughts) that serve as intermediate steps toward problem solving.
## Tree of Thoughts: Deliberate Problem Solving with Large Language Models **Shunyu Yao** **Dian Yu** **Jeffrey Zhao** **Izhak Shafran** Princeton University Google DeepMind Google DeepMind Google DeepMind **Thomas L. Griffiths** **Yuan Cao** **Karthik Narasimhan** Princeton University Google DeepMind Princeton Univ...
[ "Shunyu, Yao", "Karthik, Narasimhan", "Dian, Yu", "Jeffrey, Zhao", "Izhak, Shafran", "Thomas L., Griffiths", "Yuan, Cao" ]
2023-05-17T00:00:00
NeurIPS 2023 Oral
true
1,000
27
null
http://arxiv.org/abs/2305.10601
https://arxiv.org/abs/2305.10601
https://www.semanticscholar.org/paper/2f3822eb380b5e753a6d579f31dfc3ec4c4a0820
Measuring Mathematical Problem Solving With the MATH Dataset
Many intellectual endeavors require mathematical problem solving, but this skill remains beyond the capabilities of computers. To measure this ability in machine learning models, we introduce MATH, a new dataset of 12,500 challenging competition mathematics problems. Each problem in MATH has a full step-by-step solutio...
This work introduces MATH, a new dataset of 12,500 challenging competition mathematics problems which can be used to teach models to generate answer derivations and explanations and shows that accuracy remains relatively low, even with enormous Transformer models.
## Measuring Mathematical Problem Solving With the MATH Dataset **Dan Hendrycks** **Collin Burns** **Saurav Kadavath** **Akul Arora** **Steven Basart** UC Berkeley UC Berkeley UC Berkeley UC Berkeley UChicago **Eric Tang** **Dawn Song** **Jacob Steinhardt** UC Berkeley UC Berkeley UC Berkeley **Abstract** Many in...
[ "Dan, Hendrycks", "Collin, Burns", "Dawn, Song", "Saurav, Kadavath", "Akul, Arora", "Steven, Basart", "Eric, Tang", "Jacob, Steinhardt" ]
2021-11-08T00:00:00
NeurIPS 2021
true
935
83
null
http://arxiv.org/abs/2103.03874
https://arxiv.org/abs/2103.03874
https://www.semanticscholar.org/paper/57d1e7ac339e783898f2c3b1af55737cbeee9fc5
Self-Refine: Iterative Refinement with Self-Feedback
Like humans, large language models (LLMs) do not always generate the best output on their first try. Motivated by how humans refine their written text, we introduce Self-Refine, an approach for improving initial outputs from LLMs through iterative feedback and refinement. The main idea is to generate an initial output ...
Self-Refine is introduced, an approach for improving initial outputs from LLMs through iterative feedback and refinement that demonstrates that even state-of-the-art LLMs like GPT-4 can be further improved at test time using this simple, standalone approach.
### SELF-REFINE: Iterative Refinement with Self-Feedback **Aman Madaan[1], Niket Tandon[2], Prakhar Gupta[1], Skyler Hallinan[3], Luyu Gao[1],** **Sarah Wiegreffe[2], Uri Alon[1], Nouha Dziri[2], Shrimai Prabhumoye[4], Yiming Yang[1],** **Shashank Gupta[2], Bodhisattwa Prasad Majumder[5], Katherine Hermann[6],** **Se...
[ "Sean, Welleck", "Aman, Madaan", "Nouha, Dziri", "Luyu, Gao", "Bodhisattwa Prasad, Majumder", "Uri, Alon", "Prakhar, Gupta", "Yiming, Yang", "Skyler, Hallinan", "Sarah, Wiegreffe", "Peter, Clark", "Shrimai, Prabhumoye", "Niket, Tandon", "Shashank, Gupta", "Katherine, Hermann", "Amir, Y...
2023-05-25T00:00:00
NeurIPS 2023 Poster
true
846
32
null
http://arxiv.org/abs/2303.17651
https://arxiv.org/abs/2303.17651
https://www.semanticscholar.org/paper/3aaf6a2cbad5850ad81ab5c163599cb3d523436f
Least-to-Most Prompting Enables Complex Reasoning in Large Language Models
Chain-of-thought prompting has demonstrated remarkable performance on various natural language reasoning tasks. However, it tends to perform poorly on tasks which requires solving problems harder than the exemplars shown in the prompts. To overcome this challenge of easy-to-hard generalization, we propose a novel promp...
Experimental results on tasks related to symbolic manipulation, compositional generalization, and math reasoning reveal that least-to-most prompting is capable of generalizing to more difficult problems than those seen in the prompts.
## LEAST-TO-MOST PROMPTING ENABLES COMPLEX REASONING IN LARGE LANGUAGE MODELS **Denny Zhou[∗]** **Nathanael Sch¨arli** **Le Hou** **Jason Wei** **Nathan Scales** **Xuezhi Wang** **Dale Schuurmans** **Claire Cui** **Olivier Bousquet** **Quoc Le** **Ed Chi** Google Research, Brain Team ABSTRACT Chain-of-thought promp...
[ "Jason, Wei", "Olivier, Bousquet", "Quoc V., Le", "Xuezhi, Wang", "Le, Hou", "Nathan, Scales", "Denny, Zhou", "Dale, Schuurmans", "Nathanael, Schärli", "Ed H., Chi", "Claire, Cui" ]
2022-09-29T00:00:00
ICLR 2023 Poster
true
788
50
null
https://openreview.net/forum?id=WZH7099tgfM
https://arxiv.org/abs/2205.10625
https://www.semanticscholar.org/paper/5437e8adab596d7294124c0e798708e050e25321
Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them
BIG-Bench (Srivastava et al., 2022) is a diverse evaluation suite that focuses on tasks believed to be beyond the capabilities of current language models. Language models have already made good progress on this benchmark, with the best model in the BIG-Bench paper outperforming average reported human-rater results on 6...
This work finds that applying chain-of-thought (CoT) prompting to BBH tasks enables PaLM to surpass the average human-rater performance on 10 of the 23 tasks, and Codex to surpass it on 17 of the23 tasks.
## Challenging BIG-Bench tasks and whether chain-of-thought can solve them Mirac Suzgun[π] Nathan Scales Nathanael Schärli Sebastian Gehrmann Yi Tay Hyung Won Chung Aakanksha Chowdhery Quoc V. Le Ed H. Chi Denny Zhou Jason Wei Google Research _πStanford University_ **Abstract** BIG-Bench (Srivastava et al., 20...
[ "Yi, Tay", "Jason, Wei", "Aakanksha, Chowdhery", "Quoc V., Le", "Nathan, Scales", "Denny, Zhou", "Hyung Won, Chung", "Nathanael, Schärli", "Sebastian, Gehrmann", "Ed H., Chi", "Mirac, Suzgun" ]
2022-10-17T00:00:00
ACL 2023 Findings
false
641
21
null
https://arxiv.org/abs/2210.09261v1
https://arxiv.org/abs/2210.09261
https://www.semanticscholar.org/paper/663a41c866d49ce052801fbc88947d39764cad29
Reflexion: language agents with verbal reinforcement learning
Large language models (LLMs) have been increasingly used to interact with external environments (e.g., games, compilers, APIs) as goal-driven agents. However, it remains challenging for these language agents to quickly and efficiently learn from trial-and-error as traditional reinforcement learning methods require exte...
Reflexion is a novel framework to reinforce language agents not by updating weights, but instead through linguistic feedback, which obtains significant improvements over a baseline agent across diverse tasks (sequential decision-making, coding, language reasoning).
## Reflexion: Language Agents with Verbal Reinforcement Learning **Noah Shinn** Northeastern University ``` noahshinn024@gmail.com ``` **Federico Cassano** Northeastern University ``` cassano.f@northeastern.edu ``` **Edward Berman** Northeastern University ``` berman.ed@northeastern.edu ``` **Karthik Narasimhan*...
[ "Shunyu, Yao", "Noah, Shinn", "Karthik, Narasimhan", "Federico, Cassano", "Edward, Berman", "Ashwin, Gopinath" ]
2023-10-10T00:00:00
NeurIPS 2023 Poster
true
615
7
null
http://arxiv.org/abs/2303.11366
https://arxiv.org/abs/2303.11366
https://www.semanticscholar.org/paper/0671fd553dd670a4e820553a974bc48040ba0819
Galactica: A Large Language Model for Science
Information overload is a major obstacle to scientific progress. The explosive growth in scientific literature and data has made it ever harder to discover useful insights in a large mass of information. Today scientific knowledge is accessed through search engines, but they are unable to organize scientific knowledge ...
Galactica is introduced: a large language model that can store, combine and reason about scientific knowledge, and sets a new state-of-the-art on downstream tasks such as PubMedQA and MedMCQA dev of 77.6% and 52.9%.
## Galactica: A Large Language Model for Science **Ross Taylor** **Marcin Kardas** **Guillem Cucurull** **Thomas Scialom** **Anthony Hartshorn** **Elvis Saravia** **Andrew Poulton** **Viktor Kerkez** **Robert Stojnic** Meta AI **Abstract** Information overload is a major obstacle to scientific progress. The ex...
[ "Elvis, Saravia", "Thomas, Scialom", "Guillem, Cucurull", "Anthony, Hartshorn", "Marcin, Kardas", "Viktor, Kerkez", "Andrew, Poulton", "Ross, Taylor", "Robert, Stojnic" ]
2022-11-16T00:00:00
null
false
580
17
null
http://arxiv.org/abs/2211.09085
https://arxiv.org/abs/2211.09085
https://www.semanticscholar.org/paper/7d645a3fd276918374fd9483fd675c28e46506d1
Are NLP Models really able to Solve Simple Math Word Problems?
The problem of designing NLP solvers for math word problems (MWP) has seen sustained research activity and steady gains in the test accuracy. Since existing solvers achieve high performance on the benchmark datasets for elementary level MWPs containing one-unknown arithmetic word problems, such problems are often consi...
It is shown that MWP solvers that do not have access to the question asked in the MWP can still solve a large fraction of MWPs, and models that treat MWPs as bag-of-words can also achieve surprisingly high accuracy.
## Are NLP Models really able to Solve Simple Math Word Problems? **Arkil Patel** **Satwik Bhattamishra** **Navin Goyal** Microsoft Research India arkil.patel@gmail.com, {t-satbh,navingo}@microsoft.com **Abstract** The problem of designing NLP solvers for math word problems (MWP) has seen sustained research activit...
[ "Satwik, Bhattamishra", "Arkil, Patel", "Navin, Goyal" ]
2021-04-15T00:00:00
NAACL 2021 Main
true
574
63
null
http://arxiv.org/abs/2103.07191
https://arxiv.org/abs/2103.07191
https://www.semanticscholar.org/paper/13c4e5a6122f3fa2663f63e49537091da6532f35
Solving Quantitative Reasoning Problems with Language Models
Language models have achieved remarkable performance on a wide range of tasks that require natural language understanding. Nevertheless, state-of-the-art models have generally struggled with tasks that require quantitative reasoning, such as solving mathematics, science, and engineering questions at the college level. ...
null
### Solving Quantitative Reasoning Problems with Language Models Aitor Lewkowycz[∗], Anders Andreassen[†], David Dohan[†], Ethan Dyer[†], Henryk Michalewski[†], Vinay Ramasesh[†], Ambrose Slone, Cem Anil, Imanol Schlag, Theo Gutman-Solo, Yuhuai Wu, Behnam Neyshabur[∗], Guy Gur-Ari[∗], and Vedant Misra[∗] Google Rese...
[ "Aitor, Lewkowycz", "Anders, Andreassen", "Henryk, Michalewski", "Yuhuai, Wu", "David, Dohan", "Ethan, Dyer", "Vinay, Ramasesh", "Ambrose, Slone", "Cem, Anil", "Imanol, Schlag", "Theo, Gutman-Solo", "Behnam, Neyshabur", "Guy, Gur-Ari", "Vedant, Misra" ]
2022-06-30T00:00:00
NeurIPS 2022 Poster
true
553
46
null
http://arxiv.org/abs/2206.14858
https://arxiv.org/abs/2206.14858
https://www.semanticscholar.org/paper/ab0e3d3e4d42369de5933a3b4c237780b41c0d77
Program Induction by Rationale Generation: Learning to Solve and Explain Algebraic Word Problems
Solving algebraic word problems requires executing a series of arithmetic operations—a program—to obtain a final answer. However, since programs can be arbitrarily complicated, inducing them directly from question-answer pairs is a formidable challenge. To make this task more feasible, we solve these problems by genera...
Experimental results show that indirect supervision of program learning via answer rationales is a promising strategy for inducing arithmetic programs.
# Program Induction by Rationale Generation: Learning to Solve and Explain Algebraic Word Problems **Wang Ling[♠]** **Dani Yogatama[♠]** **Chris Dyer[♠]** **Phil Blunsom[♠♦]** _♠DeepMind_ _♦University of Oxford_ _{lingwang,dyogatama,cdyer,pblunsom}@google.com_ **Abstract** Solving algebraic word problems requires ...
[ "Wang, Ling", "Chris, Dyer", "Regina, Barzilay", "Phil, Blunsom", "Min-Yen, Kan", "Dani, Yogatama" ]
2017-07-01T00:00:00
ACL 2017 Long Papers
true
543
65
null
https://aclanthology.org/P17-1015
null
https://www.semanticscholar.org/paper/b123a0d46ad917b79c43c5ae981e03ed2458ed11
Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks
Recently, there has been significant progress in teaching language models to perform step-by-step reasoning to solve complex numerical reasoning tasks. Chain-of-thoughts prompting (CoT) is by far the state-of-art method for these tasks. CoT uses language models to perform both reasoning and computation in the multi-ste...
Under both few-shot and zero-shot settings, PoT can show an average performance gain over CoT by around 12\% across all the evaluated datasets, and by combining PoT with self-consistency decoding, can achieve SoTA performance on all math problem datasets and near-SoTAperformance on financial datasets.
## Program of Thoughts Prompting: Disentangling Computa- tion from Reasoning for Numerical Reasoning Tasks _§,¶Wenhu Chen,∗_ _[§]Xueguang Ma[∗],_ _[†]Xinyi Wang,_ _[◦]William W. Cohen_ _§University of Waterloo_ _¶Vector Institute, Toronto_ _†University of California, Santa Barabra_ _◦Google Research_ ``` {wenhuchen,x9...
[ "Xueguang, Ma", "Xinyi, Wang", "Wenhu, Chen", "William W., Cohen" ]
2023-10-22T00:00:00
null
false
524
54
null
http://arxiv.org/abs/2211.12588
https://arxiv.org/abs/2211.12588
https://www.semanticscholar.org/paper/6c943670dca38bfc7c8b477ae7c2d1fba1ad3691
Measuring and Narrowing the Compositionality Gap in Language Models
We investigate the ability of language models to perform compositional reasoning tasks where the overall solution depends on correctly composing the answers to sub-problems. We measure how often models can correctly answer all sub-problems but not generate the overall solution, a ratio we call the compositionality gap....
It is shown that the single-hop question answering performance improves faster than the multi-hop performance does, therefore the compositionality gap does not decrease, and while more powerful models memorize and recall more factual knowledge, they show no corresponding improvement in their ability to perform composit...
# Measuring and Narrowing the Compositionality Gap in Language Models **Ofir Press[1][,][2]** **Muru Zhang[1]** **Sewon Min[1][,][3]** **Ludwig Schmidt[1][,][4]** **Noah A. Smith[1][,][4]** **Mike Lewis[3]** 1Paul G. Allen School of Computer Science & Engineering, University of Washington 2MosaicML 3Meta AI Research ...
[ "Sewon, Min", "Muru, Zhang", "Ofir, Press", "Ludwig, Schmidt", "Houda, Bouamor", "Noah, Smith", "Juan, Pino", "Kalika, Bali", "Mike, Lewis" ]
2023-12-01T00:00:00
EMNLP 2023 Findings
false
407
14
null
https://aclanthology.org/2023.findings-emnlp.378
https://arxiv.org/abs/2210.03350
https://www.semanticscholar.org/paper/e070ff286709db28312e08b52b05539debe88146
Automatic Chain of Thought Prompting in Large Language Models
Large Language Models (LLMs) can carry out complex reasoning tasks by generating intermediate reasoning steps. These steps are triggered by what is called chain-of-thought (CoT) prompting, which comes in two flavors: one leverages a simple prompt like "Let’s think step by step" to facilitate step-by-step reasoning befo...
An automatic CoT prompting method that samples questions with diversity and generates reasoning chains to construct demonstrations and consistently matches or exceeds the performance of the CoT paradigm that requires manual designs of demonstrations.
## AUTOMATIC CHAIN OF THOUGHT PROMPTING ### IN LARGE LANGUAGE MODELS **Zhuosheng Zhang[†∗], Aston Zhang[‡], Mu Li[‡], Alex Smola[‡]** _†Shanghai Jiao Tong University, ‡Amazon Web Services_ **ABSTRACT** Large language models (LLMs) can perform complex reasoning by generating intermediate reasoning steps. Providing ...
[ "Zhuosheng, Zhang", "Aston, Zhang", "Mu, Li", "Alex, Smola" ]
2022-10-07T00:00:00
ICLR 2023 Poster
true
403
15
null
http://arxiv.org/abs/2210.03493
https://arxiv.org/abs/2210.03493
https://www.semanticscholar.org/paper/90350aa626bed47b02d0c162462e5b0ca82be6b2
MathQA: Towards Interpretable Math Word Problem Solving with Operation-Based Formalisms
We introduce a large-scale dataset of math word problems and an interpretable neural math problem solver by learning to map problems to their operation programs. Due to annotation challenges, current datasets in this domain have been either relatively small in scale or did not offer precise operational annotations over...
A large-scale dataset of math word problems and an interpretable neural math problem solver by learning to map problems to their operation programs and a new representation language to model operation programs corresponding to each math problem that aim to improve both the performance and the interpretability of the le...
## MathQA: Towards Interpretable Math Word Problem Solving with Operation-Based Formalisms Aida Amini[1], Saadia Gabriel[1], Shanchuan Lin[1], Rik Koncel-Kedziorski[1], Yejin Choi[1,2], and Hannaneh Hajishirzi[1,2] 1University of Washington 2Allen Institute for AI _{amini91, skgabrie, linsh, kedzior, yejin, hannaneh}...
[ "Aida, Amini", "Saadia, Gabriel", "Peter, Lin", "Hannaneh, Hajishirzi", "Rik, Koncel-Kedziorski", "Yejin, Choi" ]
2019-05-30T00:00:00
NAACL 2019 Main
true
393
42
null
http://arxiv.org/abs/1905.13319
null
https://www.semanticscholar.org/paper/eef7cfe8267954adbb4675576072a1d80ca7a3a8
Towards Reasoning in Large Language Models: A Survey
Reasoning is a fundamental aspect of human intelligence that plays a crucial role in activities such as problem solving, decision making, and critical thinking. In recent years, large language models (LLMs) have made significant progress in natural language processing, and there is observation that these models may exh...
A comprehensive overview of the current state of knowledge on reasoning in large language models, including techniques for improving and eliciting reasoning in these models, methods and benchmarks for evaluating reasoning abilities, findings and implications of previous research in this field, and suggestions on future...
## Towards Reasoning in Large Language Models: A Survey **Jie Huang** **Kevin Chen-Chuan Chang** Department of Computer Science, University of Illinois at Urbana-Champaign {jeffhj, kcchang}@illinois.edu **Abstract** Reasoning is a fundamental aspect of human intelligence that plays a crucial role in activities such...
[ "Jie, Huang", "Kevin Chen-Chuan, Chang", "Anna, Rogers", "Jordan, Boyd-Graber", "Naoaki, Okazaki" ]
2023-07-01T00:00:00
ACL 2023 Findings
false
386
10
null
https://aclanthology.org/2023.findings-acl.67
https://arxiv.org/abs/2212.10403
https://www.semanticscholar.org/paper/db4ab91d5675c37795e719e997a2827d3d83cd45
Solving General Arithmetic Word Problems
N/A
This is the first algorithmic approach that can handle arithmetic problems with multiple steps and operations, without depending on additional annotations or predefined templates, and it outperforms existing systems, achieving state of the art performance on benchmark datasets of arithmetic word problems.
# Solving General Arithmetic Word Problems **Subhro Roy** University of Illinois, Urbana Champaign sroy9@illinois.edu **Abstract** This paper presents a novel approach to automatically solving arithmetic word problems. This is the first algorithmic approach that can handle arithmetic problems with multiple steps an...
[ "Subhro, Roy", "Lluís, Màrquez", "Dan, Roth", "Jian, Su", "Chris, Callison-Burch" ]
2015-09-01T00:00:00
null
false
385
56
null
https://aclanthology.org/D15-1202
null
https://www.semanticscholar.org/paper/4c6fe6179c408e1fbb3871af13d1a8e64f766e54
Analysing Mathematical Reasoning Abilities of Neural Models
N/A
This paper conducts a comprehensive analysis of models from two broad classes of the most powerful sequence-to-sequence architectures and finds notable differences in their ability to resolve mathematical problems and generalize their knowledge.
null
[ "David, Saxton", "Felix, Hill", "Edward, Grefenstette", "Pushmeet, Kohli" ]
2019-01-01T00:00:00
ICLR 2019
true
360
29
null
https://www.semanticscholar.org/paper/afed6dc6900d3b37e528b9086661bba583d60bf6
null
https://www.semanticscholar.org/paper/afed6dc6900d3b37e528b9086661bba583d60bf6
Deep Learning for Symbolic Mathematics
Neural networks have a reputation for being better at solving statistical or approximate problems than at performing calculations or working with symbolic data. In this paper, we show that they can be surprisingly good at more elaborated tasks in mathematics, such as symbolic integration and solving differential equati...
It is shown that neural networks can be surprisingly good at more elaborated tasks in mathematics, such as symbolic integration and solving differential equations, and a syntax for representing these mathematical problems, and methods for generating large datasets that can be used to train sequence-to-sequence models.
## DEEP LEARNING FOR SYMBOLIC MATHEMATICS **Guillaume Lample[∗]** Facebook AI Research glample@fb.com **Franc¸ois Charton[∗]** Facebook AI Research fcharton@fb.com ABSTRACT Neural networks have a reputation for being better at solving statistical or approximate problems than at performing calculations or working...
[ "Guillaume, Lample", "François, Charton" ]
2019-12-02T00:00:00
ICLR 2020 Spotlight
true
357
24
null
http://arxiv.org/abs/1912.01412
null
https://www.semanticscholar.org/paper/b39eed03d345f5c244eac12fd1315d26eba77d62
Large Language Models Can Be Easily Distracted by Irrelevant Context
Large language models have achieved impressive performance on various natural language processing tasks. However, so far they have been evaluated primarily on benchmarks where all information in the input context is relevant for solving the task. In this work, we investigate the *distractibility* of large language mode...
This work investigates the distractibility of large language models, i.e., how the model problem-solving accuracy can be influenced by irrelevant context, and introduces Grade-School Math with Irrelevant Context (GSM-IC), an arithmetic reasoning dataset with irrelevant information in the problem description.
# Large Language Models Can Be Easily Distracted by Irrelevant Context **Freda Shi** [1 2 *] **Xinyun Chen** [1 *] **Kanishka Misra** [1 3] **Nathan Scales** [1] **David Dohan** [1] **Ed Chi** [1] **Nathanael Sch¨arli** [1] **Denny Zhou** [1] **Abstract** Large language models have achieved impressive performance ...
[ "Freda, Shi", "David, Dohan", "Kanishka, Misra", "Nathan, Scales", "Denny, Zhou", "Nathanael, Schärli", "Xinyun, Chen", "Ed H., Chi" ]
2023-07-03T00:00:00
ICML 2023 Poster
true
353
6
null
https://proceedings.mlr.press/v202/shi23a.html
https://arxiv.org/abs/2302.00093
https://www.semanticscholar.org/paper/3d68522abfadfc8ee6b7ec9edaaf91f1b2f38e5e
Learning to Automatically Solve Algebra Word Problems
No summary was provided.
An approach for automatically learning to solve algebra word problems by reasons across sentence boundaries to construct and solve a system of linear equations, while simultaneously recovering an alignment of the variables and numbers to the problem text.
# Learning to Automatically Solve Algebra Word Problems **Nate Kushman[†], Yoav Artzi[‡], Luke Zettlemoyer[‡], and Regina Barzilay[†]** _† Computer Science and Articial Intelligence Laboratory, Massachusetts Institute of Technology_ _{nkushman, regina}@csail.mit.edu_ _‡ Computer Science & Engineering, University of W...
[ "Yoav, Artzi", "Nate, Kushman", "Kristina, Toutanova", "Regina, Barzilay", "Luke, Zettlemoyer", "Hua, Wu" ]
2014-06-01T00:00:00
ACL 2014 Long Papers
true
347
39
null
https://aclanthology.org/P14-1026
null
https://www.semanticscholar.org/paper/3a395daf6c97c084cf9c3827384c53caf6502921
Let's Verify Step by Step
In recent years, large language models have greatly improved in their ability to perform complex multi-step reasoning. However, even state-of-the-art models still regularly produce logical mistakes. To train more reliable models, we can turn either to outcome supervision, which provides feedback for a final result, or ...
This work conducts its own investigation, finding that process supervision significantly outperforms outcome supervision for training models to solve problems from the challenging MATH dataset and shows that active learning significantly improves the efficacy of process supervision.
# Let’s Verify Step by Step **Hunter Lightman[∗]** **Vineet Kosaraju[∗]** **Yura Burda[∗]** **Harri Edwards** **Bowen Baker** **Teddy Lee** **Jan Leike** **John Schulman** **Ilya Sutskever** **Karl Cobbe[∗]** OpenAI **Abstract** In recent years, large language models have greatly improved in their ability to perf...
[ "John, Schulman", "Hunter, Lightman", "Vineet, Kosaraju", "Yura, Burda", "Bowen, Baker", "Teddy, Lee", "Harri, Edwards", "Karl, Cobbe", "Jan, Leike", "Ilya, Sutskever" ]
2023-01-01T00:00:00
ICLR 2024 Poster
true
344
28
null
https://arxiv.org/abs/2305.20050
https://arxiv.org/abs/2305.20050
https://www.semanticscholar.org/paper/be8db99310602d66bba64bcf41a572c45816fbfc
Learning to Solve Arithmetic Word Problems with Verb Categorization
N/A
The paper analyzes the arithmetic-word problems “genre”, identifying seven categories of verbs used in such problems, and reports the first learning results on this task without reliance on predefined templates and makes the data publicly available.
# Learning to Solve Arithmetic Word Problems with Verb Categorization **Mohammad Javad Hosseini[1], Hannaneh Hajishirzi[1], Oren Etzioni[2], and Nate Kushman[3]** 11University of Washington,{hosseini, hannaneh}@washington.edu, 2Allen Institute for AI, 2OrenE@allenai.org, 3Massachusetts Institute of Technology 3nkushm...
[ "Mohammad Javad, Hosseini", "Nate, Kushman", "Hannaneh, Hajishirzi", "Oren, Etzioni", "Alessandro, Moschitti", "Bo, Pang", "Walter, Daelemans" ]
2014-10-01T00:00:00
null
false
343
43
null
https://aclanthology.org/D14-1058
null
https://www.semanticscholar.org/paper/a7862e14b4c20cefd6dc4f611f8aa866fabf130b
Improving Factuality and Reasoning in Language Models through Multiagent Debate
Large language models (LLMs) have demonstrated remarkable capabilities in language generation, understanding, and few-shot learning in recent years. An extensive body of work has explored how their performance may be further improved through the tools of prompting, ranging from verification, self-consistency, or interm...
A complementary approach to improve language responses where multiple language model instances propose and debate their individual responses and reasoning processes over multiple rounds to arrive at a common final answer is presented, indicating that this approach significantly enhances mathematical and strategic reaso...
#### Improving Factuality and Reasoning in Language Models through Multiagent Debate **Yilun Du** [1] **Shuang Li** [1] **Antonio Torralba** [1] **Joshua B. Tenenbaum** [1] **Igor Mordatch** [2] **Abstract** Large language models (LLMs) have demonstrated remarkable capabilities in language generation, understanding...
[ "Shuang, Li", "Antonio, Torralba", "Igor, Mordatch", "Joshua B., Tenenbaum", "Yilun, Du" ]
2024-06-06T00:00:00
ICML 2024 Poster
true
340
9
null
https://openreview.net/forum?id=zj7YuTE4t8
https://arxiv.org/abs/2305.14325
https://www.semanticscholar.org/paper/4780d0a027c5c5a8e01d7cf697f6296880ffc945
Advancing mathematics by guiding human intuition with AI
The practice of mathematics involves discovering patterns and using these to formulate and prove conjectures, resulting in theorems. Since the 1960s, mathematicians have used computers to assist in the discovery of patterns and formulation of conjectures1, most famously in the Birch and Swinnerton-Dyer conjecture2, a M...
A framework through which machine learning can guide mathematicians in discovering new conjectures and theorems is presented and shown to yield mathematical insight on important open problems in different areas of pure mathematics.
null
[ "Alex, Davies", "Peter, Battaglia", "Nenad, Tomašev", "Lars, Buesing", "Petar, Veličković", "Sam, Blackwell", "Demis, Hassabis", "Daniel, Zheng", "Richard, Tanburn", "Charles, Blundell", "András, Juhász", "Marc, Lackenby", "Geordie, Williamson", "Pushmeet, Kohli" ]
2021-12-01T00:00:00
Nature
true
338
4
null
https://www.nature.com/articles/s41586-021-04086-x
null
https://www.semanticscholar.org/paper/c0cd4b4844c31a27a7900a754d0a91b160b00e55
AGIEval: A Human-Centric Benchmark for Evaluating Foundation Models
Assessing foundation models’ abilities for human-level tasks is crucial for Artificial General Intelligence (AGI) development.Traditional benchmarks, which rely on artificial datasets, may not accurately represent these capabilities. In this paper, we introduce AGIEval, a novel bilingual benchmark designed to assess fo...
null
## AGIEval: A Human-Centric Benchmark for Evaluating Foundation Models **Wanjun Zhong[∗], Ruixiang Cui[∗], Yiduo Guo, Yaobo Liang, Shuai Lu, Yanlin Wang** **Amin Saied, Weizhu Chen and Nan Duan** Microsoft ``` {t-wzhong, v-ruicui, v-yiduoguo, yaobo.liang, shuailu}@microsoft.com {yanlwang, Amin.Saied, wzchen, nand...
[ "Wanjun, Zhong", "Yaobo, Liang", "Ruixiang, Cui", "Nan, Duan", "Yiduo, Guo", "Yanlin, Wang", "Amin, Saied", "Weizhu, Chen", "Shuai, Lu" ]
2023-09-18T00:00:00
NAACL 2024 Findings
false
335
6
null
http://arxiv.org/abs/2304.06364
https://arxiv.org/abs/2304.06364
https://www.semanticscholar.org/paper/68c834c19cd126bbd6d25a3572d7205cfed76271
Graph of Thoughts: Solving Elaborate Problems with Large Language Models
We introduce Graph of Thoughts (GoT): a framework that advances prompting capabilities in large language models (LLMs) beyond those offered by paradigms such as Chain-of-Thought or Tree of Thoughts (ToT). The key idea and primary advantage of GoT is the ability to model the information generated by an LLM as an arbitra...
Graph of Thoughts is introduced: a framework that advances prompting capabilities in large language models (LLMs) beyond those offered by paradigms such as Chain-of-Thought or Tree of Thoughts, and is ensured that GoT is extensible with new thought transformations and thus can be used to spearhead new prompting schemes...
#### Graph of Thoughts: Solving Elaborate Problems with Large Language Models **Maciej Besta[1*], Nils Blach[1*], Ales Kubicek[1], Robert Gerstenberger[1],** **Michał Podstawski[2], Lukas Gianinazzi[1], Joanna Gajda[3], Tomasz Lehmann[3],** **Hubert Niewiadomski[3], Piotr Nyczyk[3], Torsten Hoefler[1]** 1ETH Zurich, ...
[ "Maciej, Besta", "Nils, Blach", "Ales, Kubicek", "Robert, Gerstenberger", "Michal, Podstawski", "Lukas, Gianinazzi", "Joanna, Gajda", "Tomasz, Lehmann", "Hubert, Niewiadomski", "Piotr, Nyczyk", "Torsten, Hoefler" ]
2024-03-24T00:00:00
AAAI 2024 Natural Language Processing
false
335
14
null
http://arxiv.org/abs/2308.09687
https://arxiv.org/abs/2308.09687
https://www.semanticscholar.org/paper/aade40af0d85b0b4fe15c97f6222d5c2e4d6d9b3
PAL: Program-aided Language Models
Large language models (LLMs) have demonstrated an impressive ability to perform arithmetic and symbolic reasoning tasks, when provided with a few examples at test time ("few-shot prompting"). Much of this success can be attributed to prompting methods such as "chain-of-thought", which employ LLMs for both understanding...
This paper presents Program-Aided Language models (PAL): a novel approach that uses the LLM to read natural language problems and generate programs as the intermediate reasoning steps, but offloads the solution step to a runtime such as a Python interpreter.
#### PAL: Program-aided Language Models **Luyu Gao** [* 1] **Aman Madaan** [* 1] **Shuyan Zhou** [* 1] **Uri Alon** [1] **Pengfei Liu** [1 2] **Yiming Yang** [1] **Jamie Callan** [1] **Graham Neubig** [1 2] _{luyug,amadaan,shuyanzh,ualon,pliu3,yiming,callan,gneubig}@cs.cmu.edu_ **Abstract** **1. Introduction** ...
[ "Aman, Madaan", "Luyu, Gao", "Shuyan, Zhou", "Uri, Alon", "Yiming, Yang", "Graham, Neubig", "Pengfei, Liu", "Jamie, Callan" ]
2023-01-27T00:00:00
ICML 2023 Poster
true
330
26
null
http://arxiv.org/abs/2211.10435
https://arxiv.org/abs/2211.10435
https://www.semanticscholar.org/paper/6c1e1cc1e0e1f8fd026fe517607b2d4535565fa7
Deep Neural Solver for Math Word Problems
This paper presents a deep neural solver to automatically solve math word problems. In contrast to previous statistical learning approaches, we directly translate math word problems to equation templates using a recurrent neural network (RNN) model, without sophisticated feature engineering. We further design a hybrid ...
Experiments conducted on a large dataset show that the RNN model and the hybrid model significantly outperform state-of-the-art statistical learning methods for math word problem solving.
# Deep Neural Solver for Math Word Problems **Yan Wang** **Xiaojiang Liu** **Shuming Shi** Tencent AI Lab _{brandenwang, kieranliu, shumingshi}@tencent.com_ **Problem: Dan have 2 pens, Jessica have 4** pens. How many pens do they have in total ? **Equation: x = 4+2** **Solution: 6** Table 1: A math word problem sy...
[ "Shuming, Shi", "Yan, Wang", "Sebastian, Riedel", "Xiaojiang, Liu", "Martha, Palmer", "Rebecca, Hwa" ]
2017-09-01T00:00:00
EMNLP 2017 Main
true
310
42
null
https://aclanthology.org/D17-1088
null
https://www.semanticscholar.org/paper/678fd7c48efe21434148b4b3482c2b8b3ee618fc
Complexity-Based Prompting for Multi-step Reasoning
We study the task of prompting large-scale language models to perform multi-step reasoning. Existing work shows that when prompted with a chain of thoughts (CoT), sequences of short sentences describing intermediate reasoning steps towards a final answer, large language models can generate new reasoning chains and pred...
This work proposes complexity-based prompting, a simple and effective example selection scheme for multi-step reasoning that substantially improves multi- step reasoning accuracy and achieves new state-of-the-art (SOTA) performance on three math benchmarks and two BigBenchHard tasks.
## COMPLEXITY-BASED PROMPTING FOR MULTI-STEP REASONING **Yao Fu[♠∗], Hao Peng[♣], Ashish Sabharwal[♣], Peter Clark[♣], Tushar Khot[♣]** _♠University of Edinburgh_ _♣Allen Institute for AI_ yao.fu@ed.ac.uk, haop@allenai.org, ashishs@allenai.org, peterc@allenai.org, tushark@allenai.org ABSTRACT We study the task of ...
[ "Yao, Fu", "Hao, Peng", "Tushar, Khot", "Peter, Clark", "Ashish, Sabharwal" ]
2023-01-30T00:00:00
ICLR 2023 Poster
true
309
25
null
http://arxiv.org/abs/2210.00720
https://arxiv.org/abs/2210.00720
https://www.semanticscholar.org/paper/c88cafa3e980765a64febe369ceb7c2aa7261d2a
MAWPS: A Math Word Problem Repository
No summary was provided.
MAWPS allows for the automatic construction of datasets with particular characteristics, providing tools for tuning the lexical and template overlap of a dataset as well as for filtering ungrammatical problems from web-sourced corpora.
# MAWPS: A Math Word Problem Repository **Rik Koncel-Kedziorski*[1], Subhro Roy*[2], Aida Amini[1],** **Nate Kushman[3], and Hannaneh Hajishirzi[1]** 1University of Washington, 2University of Illinois Urbana Champaign, 3Microsoft Research _{kedzior, amini91, hannaneh}@uw.edu, sroy9@illinois.edu, nkushman@microsoft.co...
[ "Subhro, Roy", "Aida, Amini", "Nate, Kushman", "Hannaneh, Hajishirzi", "Rik, Koncel-Kedziorski", "Kevin, Knight", "Ani, Nenkova", "Owen, Rambow" ]
2016-06-01T00:00:00
NAACL 2016 Main
true
308
53
null
https://aclanthology.org/N16-1136
null
https://www.semanticscholar.org/paper/2bdbb07bc12b8d8c332b7a84aa05e76218c07cd9
Transformers as Soft Reasoners over Language
Beginning with McCarthy's Advice Taker (1959), AI has pursued the goal of providing a system with explicit, general knowledge and having the system reason over that knowledge. However, expressing the knowledge in a formal (logical or probabilistic) representation has been a major obstacle to this research. This paper i...
This work trains transformers to reason (or emulate reasoning) over natural language sentences using synthetically generated data, thus bypassing a formal representation and suggesting a new role for transformers, namely as limited "soft theorem provers" operating over explicit theories in language.
## Transformers as Soft Reasoners over Language **Peter Clark, Oyvind Tafjord, Kyle Richardson** Allen Institute for AI, Seattle, WA _{peterc,oyvindt,kyler}@allenai.org_ **Abstract** Beginning with McCarthy’s Advice Taker (1959), AI has pursued the goal of providing a system with explicit, general knowledge and hav...
[ "Kyle, Richardson", "Oyvind, Tafjord", "Peter, Clark" ]
2020-05-05T00:00:00
IJCAI 2020 Natural Language Processing
false
308
9
null
http://arxiv.org/abs/2002.05867
https://arxiv.org/abs/2002.05867
https://www.semanticscholar.org/paper/9e12539d92088001e08b1e903c490127c479de4c
Mathematical Capabilities of ChatGPT
We investigate the mathematical capabilities of two iterations of ChatGPT (released 9-January-2023 and 30-January-2023) and of GPT-4 by testing them on publicly available datasets, as well as hand-crafted ones, using a novel methodology. In contrast to formal mathematics, where large databases of formal proofs are avai...
It is found that ChatGPT can be used most successfully as a mathematical assistant for querying facts, acting as a Mathematical search engine and knowledge base interface, and GPT-4 can additionally be used for undergraduate-level mathematics but fails on graduate-level difficulty.
### Mathematical Capabilities of ChatGPT Simon Frieder[∗1,4], Luca Pinchetti[1], Ryan-Rhys Griffiths[3], Tommaso Salvatori[2], Thomas Lukasiewicz[1,2], Philipp Christian Petersen[4,5], Alexis Chevalier[6], and Julius Berner[4] 1Department of Computer Science, University of Oxford, Oxford, UK 2Institute of Logic and ...
[ "Simon, Frieder", "Luca, Pinchetti", "Ryan-Rhys, Griffiths", "Tommaso, Salvatori", "Thomas, Lukasiewicz", "Philipp Christian, Petersen", "Alexis, Chevalier", "Julius, Berner" ]
2023-01-31T00:00:00
NeurIPS 2023
true
299
3
null
http://arxiv.org/abs/2301.13867
https://arxiv.org/abs/2301.13867
https://www.semanticscholar.org/paper/3f2cb353c7528efafb847309ab1e1e95245740a4
Reasoning with Language Model is Planning with World Model
Large language models (LLMs) have shown remarkable reasoning capabilities, particularly with Chain-of-Thought-style prompts. However, LLMs can still struggle with problems that are easy for humans, such as generating action plans for executing tasks or performing complex math or logical reasoning. This is due to LLMs’ ...
A new LLM reasoning framework, RAP, which repurposes the LLM as both a world model and a reasoning agent, and incorporates a principled planning algorithm (based on Monto Carlo Tree Search) for strategic exploration in the vast reasoning space.
## Reasoning with Language Model is Planning with World Model **Shibo Hao[∗♣]** **Yi Gu[∗][♣]** **Haodi Ma[♢]** **Joshua Jiahua Hong[♣]** **Zhen Wang[♣♠]** **Daisy Zhe Wang[♢]** **Zhiting Hu[♣]** _♣UC San Diego, ♢University of Florida_ _♠Mohamed bin Zayed University of Artificial Intelligence_ {s5hao, yig025, jjhon...
[ "Shibo, Hao", "Zhen, Wang", "Yi, Gu", "Haodi, Ma", "Joshua Jiahua, Hong", "Daisy Zhe, Wang", "Zhiting, Hu" ]
2023-10-23T00:00:00
EMNLP 2023 Main
true
289
9
null
http://arxiv.org/abs/2305.14992
https://arxiv.org/abs/2305.14992
https://www.semanticscholar.org/paper/5dbffedcabe3fa43060ebbe2b1789500edfd871f
ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning
Charts are very popular for analyzing data. When exploring charts, people often ask a variety of complex reasoning questions that involve several logical and arithmetic operations. They also commonly refer to visual features of a chart in their questions. However, most existing datasets do not focus on such complex rea...
This work presents two transformer-based models that combine visual features and the data table of the chart in a unified way to answer questions and achieves the state-of-the-art results on the previous datasets as well as on the benchmark.
## ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning **Ahmed Masry[♣], Do Xuan Long[♠], Jia Qing Tan[♠], Shafiq Joty[♠♦], Enamul Hoque[♣]** ♣York University, Canada ♠Nanyang Technological University, Singapore, ♦Salesforce Research ♣{masry20, enamulh}@yorku.ca ♠{xuanlong001@e....
[ "Ahmed, Masry", "Do Xuan, Long", "Shafiq, Joty", "Jia Qing, Tan", "Enamul, Hoque" ]
2022-01-01T00:00:00
ACL 2022 Findings
false
285
1
null
https://arxiv.org/abs/2203.10244
https://arxiv.org/abs/2203.10244
https://www.semanticscholar.org/paper/b611c501269224702d1a9942c8600a31ec66ab28
Decomposed Prompting: A Modular Approach for Solving Complex Tasks
Few-shot prompting is a surprisingly powerful way to use Large Language Models (LLMs) to solve various tasks. However, this approach struggles as the task complexity increases or when the individual reasoning steps of the task themselves are hard to learn, especially when embedded in more complex tasks. To address this...
It is shown that the flexibility and modularity of Decomposed Prompting allows it to outperform prior work on few-shot prompting using GPT3 and to incorporate a symbolic information retrieval within the decomposition framework, leading to improved performance on both tasks.
# Decomposed Prompting : A MODULAR APPROACH FOR SOLVING COMPLEX TASKS **Tushar Khot[♣], Harsh Trivedi[♡], Matthew Finlayson[♣], Yao Fu[♠∗],** **Kyle Richardson[♣], Peter Clark[♣], Ashish Sabharwal[♣]** _♣Allen Institute for AI_ _♡Stony Brook University_ _♠University of Edinburgh_ tushark@allenai.org, hjtrivedi@cs.sto...
[ "Matthew, Finlayson", "Yao, Fu", "Harsh, Trivedi", "Kyle, Richardson", "Tushar, Khot", "Peter, Clark", "Ashish, Sabharwal" ]
2022-09-29T00:00:00
ICLR 2023 Poster
true
284
18
null
https://openreview.net/forum?id=_nGgzQjzaRy
https://arxiv.org/abs/2210.02406
https://www.semanticscholar.org/paper/07955e96cbd778d0ae2a68f09d073b866dd84c2a
Selection-Inference: Exploiting Large Language Models for Interpretable Logical Reasoning
Large language models (LLMs) have been shown to be capable of impressive few-shot generalisation to new tasks. However, they still tend to perform poorly on multi-step logical reasoning problems. Here we carry out a comprehensive evaluation of LLMs on 46 tasks that probe different aspects of logical reasoning. We show ...
A Selection-Inference (SI) framework is proposed that exploits pre-trained LLMs as general processing modules, and alternates between selection and inference to generate a series of interpretable, casual reasoning steps leading to the final answer.
_2022-5-20_ # Selection-Inference: Exploiting Large Language Models for Interpretable Logical Reasoning **Antonia Creswell[1], Murray Shanahan[1]** **and Irina Higgins[1]** 1DeepMind **Large language models (LLMs) have been shown to be capable of impressive few-shot generalisation** **to new tasks. However, they ...
[ "Murray, Shanahan", "Antonia, Creswell", "Irina, Higgins" ]
2022-05-19T00:00:00
ICLR 2023 Notable-top-5%25
true
275
14
null
http://arxiv.org/abs/2205.09712
https://arxiv.org/abs/2205.09712
https://www.semanticscholar.org/paper/d48b29889241551e1ee6622fa78c3fa4159255dd
DVQA: Understanding Data Visualizations via Question Answering
Bar charts are an effective way to convey numeric information, but today's algorithms cannot parse them. Existing methods fail when faced with even minor variations in appearance. Here, we present DVQA, a dataset that tests many aspects of bar chart understanding in a question answering framework. Unlike visual questio...
DVQA is presented, a dataset that tests many aspects of bar chart understanding in a question answering framework and two strong baselines are proposed that perform considerably better than current VQA algorithms.
### DVQA: Understanding Data Visualizations via Question Answering Kushal Kafle[1][,][∗] Brian Price[2] Scott Cohen[2] Christopher Kanan[1] 1Rochester Institute of Technology 2Adobe Research 1{kk6055, kanan}@rit.edu 2{bprice, scohen}@adobe.com **Abstract** _Bar charts are an effective way to convey numeric in-_ ...
[ "Kushal, Kafle", "Brian, Price", "Scott, Cohen", "Christopher, Kanan" ]
2018-03-29T00:00:00
null
false
268
3
null
http://arxiv.org/abs/1801.08163
null
https://www.semanticscholar.org/paper/7289a240c9425bc7cad87b3b835e5f0cac22f488
LLM+P: Empowering Large Language Models with Optimal Planning Proficiency
Large language models (LLMs) have demonstrated remarkable zero-shot generalization abilities: state-of-the-art chatbots can provide plausible answers to many common questions that arise in daily life. However, so far, LLMs cannot reliably solve long-horizon planning problems. By contrast, classical planners, once a pro...
LLM+P is the first framework that incorporates the strengths of classical planners into large language models, and is able to provide optimal solutions for most problems, while LLMs fail to provide even feasible plans for most Problems.
## LLM+P: Empowering Large Language Models with Optimal Planning Proficiency Bo Liu[∗][†], Yuqian Jiang[∗][†], Xiaohan Zhang[‡], Qiang Liu[†], Shiqi Zhang[‡], Joydeep Biswas[†], Peter Stone[†§] **_Abstract— Large language models (LLMs) have demon-_** **strated remarkable zero-shot generalization abilities: state-** ...
[ "Bo, Liu", "Yuqian, Jiang", "Peter, Stone", "Xiaohan, Zhang", "Qiang, Liu", "Shiqi, Zhang", "Joydeep, Biswas" ]
2023-09-27T00:00:00
null
false
263
5
null
http://arxiv.org/abs/2304.11477
https://arxiv.org/abs/2304.11477
https://www.semanticscholar.org/paper/003ef1cd670d01af05afa0d3c72d72228f494432
WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct
Large language models (LLMs), such as GPT-4, have shown remarkable performance in natural language processing (NLP) tasks, including challenging mathematical reasoning. However, most existing open-source models are only pre-trained on large-scale internet data and without math-related optimization. In this paper, we pr...
WizardMath is presented, which enhances the mathematical reasoning abilities of Llama-2, by applying the proposed Reinforcement Learning from Evol-Instruct Feedback (RLEIF) method to the domain of math.
## WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct **Haipeng Luo[2][∗]** **Qingfeng Sun[1][∗]** **Can Xu[1][†]** **Pu Zhao[1]** **Jianguang Lou[1]** **Chongyang Tao[1]** **Xiubo Geng[1]** **Qingwei Lin[1]** **Shifeng Chen[2][†]** **Dongmei Zhang[1]** 1Microsoft 2S...
[ "Haipeng, Luo", "Qingfeng, Sun", "Can, Xu", "Pu, Zhao", "Jianguang, Lou", "Dongmei, Zhang", "Chongyang, Tao", "Xiubo, Geng", "Qingwei, Lin", "Shifeng, Chen" ]
2023-08-18T00:00:00
ICLR 2025 Submission
false
262
27
null
http://arxiv.org/abs/2308.09583
https://arxiv.org/abs/2308.09583
https://www.semanticscholar.org/paper/dd18782960f9ee4c66b79e1518b342ad3f8d19e7
STaR: Bootstrapping Reasoning With Reasoning
Generating step-by-step "chain-of-thought" rationales improves language model performance on complex reasoning tasks like mathematics or commonsense question-answering. However, inducing language model rationale generation currently requires either constructing massive rationale datasets or sacrificing accuracy by usin...
A technique to iteratively leverage a small number of rationale examples and a large dataset without rationales to bootstrap the ability to perform successively more complex reasoning, called STaR, which lets a model improve itself by learning from its own generated reasoning.
## STaR: Self-Taught Reasoner ### Bootstrapping Reasoning With Reasoning **Eric Zelikman[∗][1], Yuhuai Wu[∗][12], Jesse Mu[1], Noah D. Goodman[1]** 1Department of Computer Science, Stanford University 2 Google Research ``` {ezelikman, yuhuai, muj, ngoodman}@stanford.edu ``` **Abstract** Generating step-by-step "c...
[ "Eric, Zelikman", "Yuhuai, Wu", "Jesse, Mu", "Noah D., Goodman" ]
2022-05-20T00:00:00
NeurIPS 2022 Poster
true
260
20
null
http://arxiv.org/abs/2203.14465
https://arxiv.org/abs/2203.14465
https://www.semanticscholar.org/paper/23dd78e424d32f6a48660dcd67ce994b8a7db8be
A Diverse Corpus for Evaluating and Developing English Math Word Problem Solvers
We present ASDiv (Academia Sinica Diverse MWP Dataset), a diverse (in terms of both language patterns and problem types) English math word problem (MWP) corpus for evaluating the capability of various MWP solvers. Existing MWP corpora for studying AI progress remain limited either in language usage patterns or in probl...
A metric to measure the lexicon usage diversity of a given MWP corpus is proposed, and it is demonstrated that ASDiv (Academia Sinica Diverse MWP Dataset) is more diverse than existing corpora.
# A Diverse Corpus for Evaluating and Developing English Math Word Problem Solvers **Shen-Yun Miao, Chao-Chun Liang, and Keh-Yih Su** Institute of Information Science, Academia Sinica, Taiwan miao.iisas@gmail.com, {ccliang, kysu}@iis.sinica.edu.tw **Abstract** We present _ASDiv (Academia_ **Sinica** **Di-** **verse...
[ "Shen-Yun, Miao", "Chao-Chun, Liang", "Keh-Yih, Su" ]
2021-06-29T00:00:00
ACL 2020
true
253
37
null
http://arxiv.org/abs/2106.15772
https://arxiv.org/abs/2106.15772
https://www.semanticscholar.org/paper/f13e41d24e5d0a68ca662c1b49de398a6fb68251
Parsing Algebraic Word Problems into Equations
This paper formalizes the problem of solving multi-sentence algebraic word problems as that of generating and scoring equation trees. We use integer linear programming to generate equation trees and score their likelihood by learning local and global discriminative models. These models are trained on a small set of wor...
null
# Parsing Algebraic Word Problems into Equations **Rik Koncel-Kedziorski, Hannaneh Hajishirzi,** **Ashish Sabharwal[†], Oren Etzioni[†], and Siena Dumas Ang** University of Washington, _[†]Allen Institute for AI_ _{kedzior,hannaneh,sienaang}@uw.edu, {ashishs,orene}@allenai.org_ **Abstract** This paper formalizes th...
[ "Hannaneh, Hajishirzi", "Oren, Etzioni", "Rik, Koncel-Kedziorski", "Siena Dumas, Ang", "Michael, Collins", "Lillian, Lee", "Ashish, Sabharwal" ]
2015-01-01T00:00:00
null
false
252
37
null
https://aclanthology.org/Q15-1042
null
https://www.semanticscholar.org/paper/17230f5b3956188055a48c5f4f61d131cce0662f
Large Language Models Cannot Self-Correct Reasoning Yet
Large Language Models (LLMs) have emerged as a groundbreaking technology with their unparalleled text generation capabilities across various applications. Nevertheless, concerns persist regarding the accuracy and appropriateness of their generated content. A contemporary methodology, self-correction, has been proposed ...
It is indicated that LLMs struggle to self-correct their responses without external feedback, and at times, their performance even degrades after self-correction.
# LARGE LANGUAGE MODELS CANNOT SELF-CORRECT REASONING YET **Jie Huang[1][,][2][∗]** **Xinyun Chen[1][∗]** **Swaroop Mishra[1]** **Huaixiu Steven Zheng[1]** **Adams Wei Yu[1]** **Xinying Song[1]** **Denny Zhou[1]** 1Google DeepMind 2University of Illinois at Urbana-Champaign jeffhj@illinois.edu, {xinyunchen, dennyzh...
[ "Swaroop, Mishra", "Jie, Huang", "Xinying, Song", "Adams Wei, Yu", "Denny, Zhou", "Huaixiu Steven, Zheng", "Xinyun, Chen" ]
2023-10-13T00:00:00
ICLR 2024 Poster
true
236
14
null
https://openreview.net/forum?id=IkmD3fKBPQ
https://arxiv.org/abs/2310.01798
https://www.semanticscholar.org/paper/6d4bacb69923e1e94fb4de468b939ce6db32fb51
Generative Language Modeling for Automated Theorem Proving
We explore the application of transformer-based language models to automated theorem proving. This work is motivated by the possibility that a major limitation of automated theorem provers compared to humans -- the generation of original mathematical terms -- might be addressable via generation from language models. We...
This work presents an automated prover and proof assistant, GPT-f, for the Metamath formalization language, and analyzes its performance, finding new short proofs that were accepted into the mainMetamath library, which is to this knowledge, the first time a deep-learning based system has contributed proofs that are ado...
## Generative Language Modeling for Automated Theorem Proving **Stanislas Polu** OpenAI ``` spolu@openai.com ``` **Ilya Sutskever** OpenAI ``` ilyasu@openai.com ``` **Abstract** We explore the application of transformer-based language models to automated theorem proving. This work is motivated by the possibilit...
[ "Stanislas, Polu", "Ilya, Sutskever" ]
2020-09-01T00:00:00
N/A
false
234
40
null
https://arxiv.org/abs/2009.03393
https://arxiv.org/abs/2009.03393
https://www.semanticscholar.org/paper/5fe0a4af3bd1479d5e39fbda2215c86bce54722b
FigureQA: An Annotated Figure Dataset for Visual Reasoning
We introduce FigureQA, a visual reasoning corpus of over one million question-answer pairs grounded in over 100,000 images. The images are synthetic, scientific-style figures from five classes: line plots, dot-line plots, vertical and horizontal bar graphs, and pie charts. We formulate our reasoning task by generating ...
FigureQA is envisioned as a first step towards developing models that can intuitively recognize patterns from visual representations of data, and preliminary results indicate that the task poses a significant machine learning challenge.
## FIGUREQA: AN ANNOTATED FIGURE DATASET FOR VISUAL REASONING **Samira Ebrahimi Kahou[1][∗], Vincent Michalski[2][∗†], Adam Atkinson[1],** **Ákos Kádár[3][†], Adam Trischler[1], Yoshua Bengio[3]** 1Microsoft Research Montréal 2Université de Montréal, MILA 3Tilburg University ABSTRACT We introduce FigureQA, a visua...
[ "Yoshua, Bengio", "Vincent, Michalski", "Samira Ebrahimi, Kahou", "Adam, Atkinson", "Akos, Kadar", "Adam, Trischler" ]
2018-02-22T00:00:00
ICLR 2018
true
233
2
null
http://arxiv.org/abs/1710.07300
null
https://www.semanticscholar.org/paper/cb6be69c67b0b15ebbda89a126f4dd62a4d32958
Large Language Models Are Reasoning Teachers
Recent works have shown that chain-of-thought (CoT) prompting can elicit language models to solve complex reasoning tasks, step-by-step. However, prompt-based CoT methods are dependent on very large models such as GPT-3 175B which are prohibitive to deploy at scale. In this paper, we use these large models as reasoning...
This paper uses very large models as reasoning teachers to enable complex reasoning in smaller models and reduce model size requirements by several orders of magnitude, and proposes Fine-tune-CoT, a method that generates reasoning samples from very large teacher models to fine-tunes smaller models.
### Large Language Models Are Reasoning Teachers **Namgyu Ho, Laura Schmid, Se-Young Yun** KAIST {itsnamgyu, laura.schmid, yunseyoung}@kaist.ac.kr Questions ## ? ? ? Reasoning samples **Abstract** Recent works have shown that chain-of-thought (CoT) prompting can elicit language models to solve complex reasonin...
[ "Namgyu, Ho", "Laura, Schmid", "Se-Young, Yun", "Anna, Rogers", "Jordan, Boyd-Graber", "Naoaki, Okazaki" ]
2023-07-01T00:00:00
ACL 2023 Long Papers
true
232
15
null
https://aclanthology.org/2023.acl-long.830
https://arxiv.org/abs/2212.10071
https://www.semanticscholar.org/paper/a9e3e5dd7b30890553b7ae1c41f932e99192bb44
MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuning
We introduce MAmmoTH, a series of open-source large language models (LLMs) specifically tailored for general math problem-solving. The MAmmoTH models are trained on MathInstruct, our meticulously curated instruction tuning dataset. MathInstruct is compiled from 13 math datasets with intermediate rationales, six of whic...
The MAmmoTH series substantially outperform existing open-source models on nine mathematical reasoning datasets across all scales with an average accuracy gain between 16% and 32%, and underscores the importance of diverse problem coverage and the use of hybrid rationales in developing superior math generalist models.
## MAMMOTH: BUILDING MATH GENERALIST MODELS ### THROUGH HYBRID INSTRUCTION TUNING **Xiang Yue** **, ‡Xingwei Qu, †Ge Zhang,** **Yao Fu, §Wenhao Huang,** ♣Huan Sun,∗ **Yu Su, †Wenhu Chen** - † ♣University of Waterloo, ♣ The Ohio State University,∗ ‡HKUST, University of Edinburgh, §01.AI yue.149@osu.edu, wenhuchen@u...
[ "Xiang, Yue", "Yao, Fu", "Xingwei, Qu", "Ge, Zhang", "Wenhu, Chen", "Wenhao, Huang", "Huan, Sun", "Yu, Su" ]
2023-10-02T00:00:00
ICLR 2024 Spotlight
true
232
24
null
http://arxiv.org/abs/2309.05653
https://arxiv.org/abs/2309.05653
https://www.semanticscholar.org/paper/a3dd7d33dfaa9e02e43d92e900cba01f52d8c4b9
Reasoning with Language Model Prompting: A Survey
Reasoning, as an essential ability for complex problem-solving, can provide back-end support for various real-world applications, such as medical diagnosis, negotiation, etc. This paper provides a comprehensive survey of cutting-edge research on reasoning with language model prompting. We introduce research works with ...
This paper provides a comprehensive survey of cutting-edge research on reasoning with language model prompting with comparisons and summaries and provides systematic resources to help beginners.
## Reasoning with Language Model Prompting: A Survey **Shuofei Qiao[1][∗], Yixin Ou[1][∗], Ningyu Zhang[1][†], Xiang Chen[1], Yunzhi Yao[1],** **Shumin Deng[4], Chuanqi Tan[3], Fei Huang[3], Huajun Chen[1][,][2][†]** 1 Zhejiang University & AZFT Joint Lab for Knowledge Engine 2 Donghai Laboratory 3 Alibaba Group 4 Na...
[ "Xiang, Chen", "Shuofei, Qiao", "Yixin, Ou", "Ningyu, Zhang", "Shumin, Deng", "Yunzhi, Yao", "Chuanqi, Tan", "Fei, Huang", "Huajun, Chen" ]
2023-09-18T00:00:00
ACL 2023 Long Papers
true
220
5
null
http://arxiv.org/abs/2212.09597
https://arxiv.org/abs/2212.09597
https://www.semanticscholar.org/paper/6845bea94b2fb17d4377b3bb2bd10f73a959f9cc
CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing
Recent developments in large language models (LLMs) have been impressive. However, these models sometimes show inconsistencies and problematic behavior, such as hallucinating facts, generating flawed code, or creating offensive and toxic content. Unlike these models, humans typically utilize external tools to cross-che...
A framework called CRITIC is introduced that allows LLMs, which are essentially"black boxes" to validate and progressively amend their own outputs in a manner similar to human interaction with tools.
## CRITIC: LARGE LANGUAGE MODELS CAN SELF- CORRECT WITH TOOL-INTERACTIVE CRITIQUING **Zhibin Gou[12][∗], Zhihong Shao[12][∗], Yeyun Gong[2], Yelong Shen[3],** **Yujiu Yang[1][†], Nan Duan[2], Weizhu Chen[3]** 1Tsinghua University 2Microsoft Research Asia, 3Microsoft Azure AI {gzb22,szh19}@mails.tsinghua.edu.cn, yang....
[ "Zhibin, Gou", "Zhihong, Shao", "Nan, Duan", "Yeyun, Gong", "Yelong, Shen", "Yujiu, Yang", "Weizhu, Chen" ]
2024-02-21T00:00:00
ICLR 2024 Poster
true
215
8
null
http://arxiv.org/abs/2305.11738
https://arxiv.org/abs/2305.11738
https://www.semanticscholar.org/paper/bcdaf6c98ddbd6809cf6241aa77200d7394db163
Language models are multilingual chain-of-thought reasoners
We evaluate the reasoning abilities of large language models in multilingual settings. We introduce the Multilingual Grade School Math (MGSM) benchmark, by manually translating 250 grade-school math problems from the GSM8K dataset (Cobbe et al., 2021) into ten typologically diverse languages. We find that the ability t...
It is found that the ability to solve MGSM problems via chain-of-thought prompting emerges with increasing model scale, and that models have strikingly strong multilingual reasoning abilities, even in underrepresented languages such as Bengali and Swahili.
## LANGUAGE MODELS ARE MULTILINGUAL CHAIN-OF-THOUGHT REASONERS **Freda Shi[1][,][2][,][∗]** **Mirac Suzgun[1][,][3][,][∗]** **Markus Freitag[1]** **Xuezhi Wang[1]** **Suraj Srivats[4]** **Soroush Vosoughi[4]** **Hyung Won Chung[1]** **Yi Tay[1]** **Sebastian Ruder[1]** **Denny Zhou[1]** **Dipanjan Das[1]** **Jason ...
[ "Yi, Tay", "Jason, Wei", "Freda, Shi", "Xuezhi, Wang", "Markus, Freitag", "Denny, Zhou", "Hyung Won, Chung", "Suraj, Srivats", "Soroush, Vosoughi", "Sebastian, Ruder", "Dipanjan, Das", "Mirac, Suzgun" ]
2022-10-06T00:00:00
ICLR 2023 Poster
true
215
6
null
http://arxiv.org/abs/2210.03057
https://arxiv.org/abs/2210.03057
https://www.semanticscholar.org/paper/62f0db3a5ad5c795ec18fc7a6e7b01836809df57
DeepMath - Deep Sequence Models for Premise Selection
We study the effectiveness of neural sequence models for premise selection in automated theorem proving, one of the main bottlenecks in the formalization of mathematics. We propose a two stage approach for this task that yields good results for the premise selection task on the Mizar corpus while avoiding the hand-engi...
A two stage approach is proposed that yields good results for the premise selection task on the Mizar corpus while avoiding the hand-engineered features of existing state-of-the-art models.
## DeepMath - Deep Sequence Models for Premise Selection **Alexander A. Alemi** _[∗]_ Google Inc. ``` alemi@google.com ``` **Geoffrey Irving** _[∗]_ Google Inc. ``` geoffreyi@google.com ``` **François Chollet** _[∗]_ Google Inc. ``` fchollet@google.com ``` **Niklas Een** _[∗]_ Google Inc. ``` een@google.com ...
[ "Alex A., Alemi", "Josef, Urban", "Francois, Chollet", "Niklas, Een", "Geoffrey, Irving", "Christian, Szegedy" ]
2017-01-26T00:00:00
NeurIPS 2016
true
212
26
[ "Mizar" ]
http://arxiv.org/abs/1606.04442
null
https://www.semanticscholar.org/paper/f8909e8b359c72bc0e3257dadcc48a8787bca74b
MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts
Although Large Language Models (LLMs) and Large Multimodal Models (LMMs) exhibit impressive skills in various domains, their ability for mathematical reasoning within visual contexts has not been formally examined. Equipping LLMs and LMMs with this capability is vital for general-purpose AI assistants and showcases pro...
The in-depth analysis reveals that the superiority of GPT-4V is mainly attributed to its enhanced visual perception and mathematical reasoning, but it still falls short of human performance by 10.4%, which underscores the critical role that MathVista will play in the development of general-purpose AI agents capable of ...
**MATHVISTA: Evaluating Mathematical Reasoning of** **Foundation Models in Visual Contexts** **Pan Lu[1][,][3], Hritik Bansal[1], Tony Xia[1], Jiacheng Liu[2], Chunyuan Li[3],** **Hannaneh Hajishirzi[2], Hao Cheng[3], Kai-Wei Chang[1], Michel Galley[3], Jianfeng Gao[3]** 1UCLA, 2University of Washington, 3Microsoft R...
[ "Pan, Lu", "Jiacheng, Liu", "Hritik, Bansal", "Tony, Xia", "Hannaneh, Hajishirzi", "Jianfeng, Gao", "Kai-Wei, Chang", "Chunyuan, Li", "Hao, Cheng", "Michel, Galley" ]
2023-10-28T00:00:00
ICLR 2024 Oral
true
212
4
null
https://openreview.net/forum?id=hrI14X0Ltk
https://arxiv.org/abs/2310.02255
https://www.semanticscholar.org/paper/8946891e94831adc8cddb0d32311cce2445c96d2
Injecting Numerical Reasoning Skills into Language Models
Large pre-trained language models (LMs) are known to encode substantial amounts of linguistic information. However, high-level reasoning skills, such as numerical reasoning, are difficult to learn from a language-modeling objective only. Consequently, existing models for numerical reasoning have used specialized archit...
This work shows that numerical reasoning is amenable to automatic data generation, and thus one can inject this skill into pre-trained LMs, by generating large amounts of data, and training in a multi-task setup.
## Injecting Numerical Reasoning Skills into Language Models **Mor Geva[∗]** Tel Aviv University, Allen Institute for AI morgeva@mail.tau.ac.il **Ankit Gupta[∗]** Tel Aviv University ankitgupta.iitkanpur@gmail.com **Jonathan Berant** Tel Aviv University, Allen Institute for AI joberant@cs.tau.ac.il **Abstract...
[ "Mor, Geva", "Ankit, Gupta", "Jonathan, Berant" ]
2020-04-09T00:00:00
ACL 2020 Main
true
205
10
null
http://arxiv.org/abs/2004.04487
https://arxiv.org/abs/2004.04487
https://www.semanticscholar.org/paper/3dd61d97827e3f380bf9304101149a3f865051fc
Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate
Modern large language models (LLMs) like ChatGPT have shown remarkable performance on general language tasks but still struggle on complex reasoning tasks, which drives the research on cognitive behaviors of LLMs to explore human-like problem-solving strategies. Along this direction, one representative strategy is self...
A Multi-Agent Debate (MAD) framework is proposed, in which multiple agents express their arguments in the state of"tit for tat"and a judge manages the debate process to obtain a final solution.
## Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate **Tian Liang[1][*]** **Zhiwei He[2][*]** **Wenxiang Jiao[3][*]** **Xing Wang[3][†]** **Rui Wang[2]** **Yujiu Yang[1][†]** **Shuming Shi[3]** **Zhaopeng Tu[3]** 1Tsinghua University 2Shanghai Jiao Tong University 3Tencent AI Lab {li...
[ "Shuming, Shi", "Zhiwei, He", "Rui, Wang", "Tian, Liang", "Wenxiang, Jiao", "Xing, Wang", "Yujiu, Yang", "Zhaopeng, Tu" ]
2024-07-17T00:00:00
null
false
204
8
null
http://arxiv.org/abs/2305.19118
https://arxiv.org/abs/2305.19118
https://www.semanticscholar.org/paper/385c74957858e7d6856d48e72b5a902b4c1aa28c
TAT-QA: A Question Answering Benchmark on a Hybrid of Tabular and Textual Content in Finance
Hybrid data combining both tabular and textual content (e.g., financial reports) are quite pervasive in the real world. However, Question Answering (QA) over such hybrid data is largely neglected in existing research. In this work, we extract samples from real financial reports to build a new large-scale QA dataset con...
This work extracts samples from real financial reports to build a new large-scale QA dataset containing both Tabular And Textual data, named TAT-QA, where numerical reasoning is usually required to infer the answer, such as addition, subtraction, multiplication, division, counting, comparison/sorting, and the compositi...
## TAT-QA: A Question Answering Benchmark on a Hybrid of Tabular and Textual Content in Finance **Fengbin Zhu[1,2], Wenqiang Lei[1][∗], Youcheng Huang** **[3], Chao Wang[2], Shuo Zhang[4],** **Jiancheng Lv[3], Fuli Feng[1], Tat-Seng Chua[1]** 1National University of Singapore, 26Estates Pte Ltd, 3Sichuan University, ...
[ "Fengbin, Zhu", "Shuo, Zhang", "Wenqiang, Lei", "Youcheng, Huang", "Chao, Wang", "Jiancheng, Lv", "Fuli, Feng", "Tat-Seng, Chua" ]
2021-06-01T00:00:00
ACL 2021 Long Papers
true
200
10
null
http://arxiv.org/abs/2105.07624
https://arxiv.org/abs/2105.07624
https://www.semanticscholar.org/paper/b3213c84a6ff7a2f11099de783c93166e4fc02a4
A Goal-Driven Tree-Structured Neural Model for Math Word Problems
Most existing neural models for math word problems exploit Seq2Seq model to generate solution expressions sequentially from left to right, whose results are far from satisfactory due to the lack of goal-driven mechanism commonly seen in human problem solving. This paper proposes a tree-structured neural model to genera...
A treestructured neural model to generate expression tree in a goal-driven manner is proposed and experimental results on the dataset Math23K have shown that the model outperforms significantly several state-of-the-art models.
null
[ "Zhipeng, Xie", "Shichao, Sun" ]
2019-08-01T00:00:00
IJCAI 2019 Natural Language Processing
false
194
27
null
https://www.ijcai.org/proceedings/2019/736
null
https://www.semanticscholar.org/paper/5bf6ca86a4f21d07014ce60f3cd09345ad3414e2
Plan-and-Solve Prompting: Improving Zero-Shot Chain-of-Thought Reasoning by Large Language Models
Large language models (LLMs) have recently been shown to deliver impressive performance in various NLP tasks. To tackle multi-step reasoning tasks, Few-shot chain-of-thought (CoT) prompting includes a few manually crafted step-by-step reasoning demonstrations which enable LLMs to explicitly generate reasoning steps and...
The experimental results over GPT-3 show that the proposed zero-shot prompting consistently outperforms Zero- shot-CoT across all datasets by a large margin, is comparable to or exceeds Zero-shot-Program-of-Thought Prompting, and has comparable performance with 8-shot CoT prompting on the math reasoning problem.
# Plan-and-Solve Prompting: Improving Zero-Shot Chain-of-Thought Reasoning by Large Language Models **Lei Wang[1]** **Wanyu Xu[2]** **Yihuai Lan Zhiqiang Hu[3]** **Yunshi Lan[4]** **Roy Ka-Wei Lee[3]** **Ee-Peng Lim[1][∗]** 1Singapore Management University 2Southwest Jiaotong University 3Singapore University of Tech...
[ "Yunshi, Lan", "Zhiqiang, Hu", "Wanyu, Xu", "Lei, Wang", "Yihuai, Lan", "Ee-Peng, Lim", "Anna, Rogers", "Roy Ka-Wei, Lee", "Jordan, Boyd-Graber", "Naoaki, Okazaki" ]
2023-07-01T00:00:00
ACL 2023 Long Papers
true
194
5
null
https://aclanthology.org/2023.acl-long.147
https://arxiv.org/abs/2305.04091
https://www.semanticscholar.org/paper/62176de125738e3b95850d1227bac81fd646b78e
Dynamic Prompt Learning via Policy Gradient for Semi-structured Mathematical Reasoning
Mathematical reasoning, a core ability of human intelligence, presents unique challenges for machines in abstract thinking and logical reasoning. Recent large pre-trained language models such as GPT-3 have achieved remarkable progress on mathematical reasoning tasks written in text form, such as math word problems (MWP...
A novel approach is proposed, PromptPG, which utilizes policy gradient to learn to select in-context examples from a small amount of training data and then constructs the corresponding prompt for the test example, which verifies its effectiveness in selecting in- context examples.
## DYNAMIC PROMPT LEARNING VIA POLICY GRADIENT ### FOR SEMI-STRUCTURED MATHEMATICAL REASONING **Pan Lu[1][,][3], Liang Qiu[1], Kai-Wei Chang[1], Ying Nian Wu[1], Song-Chun Zhu[1],** **Tanmay Rajpurohit[2], Peter Clark[3], Ashwin Kalyan[3]** 1University of California, Los Angeles, 2Georgia Institute of Technology, 3Al...
[ "Pan, Lu", "Liang, Qiu", "Ying Nian, Wu", "Kai-Wei, Chang", "Song-Chun, Zhu", "Tanmay, Rajpurohit", "Peter, Clark", "Ashwin, Kalyan" ]
2023-03-02T00:00:00
ICLR 2023
true
181
17
null
http://arxiv.org/abs/2209.14610
https://arxiv.org/abs/2209.14610
https://www.semanticscholar.org/paper/3e565c544a8639cc9c7568833e484d7610f5e5d4
Specializing Smaller Language Models towards Multi-Step Reasoning
The surprising ability of Large Language Models (LLMs) to perform well on complex reasoning with only few-shot chain-of-thought prompts is believed to emerge only in very large-scale models. We show that such abilities can, in fact, be distilled down from GPT-3.5 (≥ 175B) to T5 variants (≤ 11B). We propose model specia...
This work shows two important aspects of model abilities: there exists a very complex balance/ tradeoff between language models' multi-dimensional abilities, and by paying the price of decreased generic ability, it can clearly lift up the scaling curve of models smaller than 10B towards a specialized multi-step math re...
# Specializing Smaller Language Models towards Multi-Step Reasoning **Yao Fu** [1] **Hao Peng** [2] **Litu Ou** [1] **Ashish Sabharwal** [2] **Tushar Khot** [2] **Abstract** The surprising ability of Large Language Models (LLMs) to perform well on complex reasoning with only few-shot chain-of-thought prompts is bel...
[ "Yao, Fu", "Hao, Peng", "Litu, Ou", "Tushar, Khot", "Ashish, Sabharwal" ]
2023-07-03T00:00:00
ICML 2023 Oral
true
177
14
null
https://proceedings.mlr.press/v202/fu23d.html
https://arxiv.org/abs/2301.12726
https://www.semanticscholar.org/paper/fbd49b25bdab98c171af49962a41139c73dacbde
Teaching Small Language Models to Reason
Chain of thought prompting successfully improves the reasoning capabilities of large language models, achieving state of the art results on a range of datasets. However, these reasoning capabilities only appear to emerge in models with at least tens of billions of parameters. In this paper, we explore the transfer of s...
This paper finetune a student model on the chain of thought outputs generated by a larger teacher model, and shows that the proposed method improves task performance across arithmetic, commonsense and symbolic reasoning datasets.
# Teaching Small Language Models to Reason **Lucie Charlotte Magister[∗]** University of Cambridge lcm67@cam.ac.uk **Jonathan Mallinson** Google Research jonmall@google.com **Jakub Adamek** Google Research enkait@google.com **Eric Malmi** Google Research emalmi@google.com **Abstract** Chain of thought prompt...
[ "Jonathan, Mallinson", "Jakub, Adamek", "Eric, Malmi", "Lucie Charlotte, Magister", "Aliaksei, Severyn", "Anna, Rogers", "Jordan, Boyd-Graber", "Naoaki, Okazaki" ]
2023-07-01T00:00:00
ACL 2023 Short Papers
false
175
15
null
https://aclanthology.org/2023.acl-short.151
https://arxiv.org/abs/2212.08410
https://www.semanticscholar.org/paper/126a4776ff8315fd506766cb8f3c722cf746ad9e
Graph-to-Tree Learning for Solving Math Word Problems
While the recent tree-based neural models have demonstrated promising results in generating solution expression for the math word problem (MWP), most of these models do not capture the relationships and order information among the quantities well. This results in poor quantity representations and incorrect solution exp...
Graph2Tree is proposed, a novel deep learning architecture that combines the merits of the graph-based encoder and tree-based decoder to generate better solution expressions for math word problem (MWP) solution expressions.
# Graph-to-Tree Learning for Solving Math Word Problems **Jipeng Zhang[1][,][2][,][∗], Lei Wang[2][,][∗], Roy Ka-Wei Lee[3], Yi Bin[1], Yan Wang[4], Jie Shao[1][,][5][,][#],** **Ee-Peng Lim[2]** 1Center for Future Media, University of Electronic Science and Technology of China 2School of Information Systems, Singapor...
[ "Yi, Bin", "Jipeng, Zhang", "Lei, Wang", "Yan, Wang", "Ee-Peng, Lim", "Jie, Shao", "Dan, Jurafsky", "Joyce, Chai", "Roy Ka-Wei, Lee", "Natalie, Schluter", "Joel, Tetreault" ]
2020-07-01T00:00:00
ACL 2020 Main
true
166
20
null
https://aclanthology.org/2020.acl-main.362
null
https://www.semanticscholar.org/paper/ac9173e4b8b34dc24c18980ee32be6bf15e7b38d
Llemma: An Open Language Model for Mathematics
We present Llemma, a large language model for mathematics. We continue pretraining Code Llama on the Proof-Pile-2, a mixture of scientific papers, web data containing mathematics, and mathematical code, yielding Llemma. On the MATH benchmark Llemma outperforms all known openly released models, as well as the unreleased...
Llemma is a large language model for mathematics that outperforms all known open base models, as well as the unreleased Minerva model suite on an equi-parameter basis, and is capable of tool use and formal theorem proving without any further finetuning.
## LLEMMA: AN OPEN LANGUAGE MODEL FOR ### MATHEMATICS **Zhangir Azerbayev** [1][,][2] **Hailey Schoelkopf** [2] **Keiran Paster** [3][,][4] **Marco Dos Santos** [5] **Stephen McAleer** [6] **Albert Q. Jiang** [5] **Jia Deng** [1] **Stella Biderman** [2] **Sean Welleck** [6][,][7] 1 2 3 4 Princeton University Eleu...
[ "Sean, Welleck", "Zhangir, Azerbayev", "Jia, Deng", "Albert Q., Jiang", "Stephen, McAleer", "Hailey, Schoelkopf", "Keiran, Paster", "Marco Dos, Santos", "Stella, Biderman" ]
2023-10-16T00:00:00
ICLR 2024
true
164
12
[ "Lean", "Isabelle" ]
http://arxiv.org/abs/2310.10631
https://arxiv.org/abs/2310.10631
https://www.semanticscholar.org/paper/b16c7d45183b9d595ab64301be019741b1528860
MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Models
Large language models (LLMs) have pushed the limits of natural language understanding and exhibited excellent problem-solving ability. Despite the great success, most existing open-source LLMs (\eg, LLaMA-2) are still far away from satisfactory for solving mathematical problems due to the complex reasoning procedures. ...
Experimental results on two popular benchmarks for mathematical reasoning demonstrate that MetaMath outperforms a suite of open-source LLMs by a significant margin.
## METAMATH: BOOTSTRAP YOUR OWN MATHEMATICAL QUESTIONS FOR LARGE LANGUAGE MODELS **Longhui Yu[1,*]** **Weisen Jiang[2,3,*]** **Han Shi[4,†]** **Jincheng Yu[3,4]** **Zhengying Liu[4]** **Yu Zhang[2]** **James T. Kwok[3]** **Zhenguo Li[4]** **Adrian Weller[1,5]** **Weiyang Liu[1,6,†]** 1University of Cambridge 2South...
[ "Longhui, Yu", "Yu, Zhang", "Weisen, Jiang", "James T., Kwok", "Zhengying, Liu", "Jincheng, Yu", "Zhenguo, Li", "Han, Shi", "Weiyang, Liu", "Adrian, Weller" ]
2023-10-09T00:00:00
ICLR 2024 Spotlight
true
163
13
null
http://arxiv.org/abs/2309.12284
https://arxiv.org/abs/2309.12284
https://www.semanticscholar.org/paper/77b1f1c6d1658d120456b9046667cf009ceb39ce
Solving olympiad geometry without human demonstrations
Proving mathematical theorems at the olympiad level represents a notable milestone in human-level automated reasoning1–4, owing to their reputed difficulty among the world’s best talents in pre-university mathematics. Current machine-learning approaches, however, are not applicable to most mathematical domains owing to...
AlphaGeometry is a neuro-symbolic system that uses a neural language model that uses a neural language model, trained from scratch on large-scale synthetic data, to guide a symbolic deduction engine through infinite branching points in challenging problems.
null
[ "Trieu H., Trinh", "Quoc V., Le", "Yuhuai, Wu", "He, He", "Thang, Luong" ]
2024-01-01T00:00:00
Nature
true
162
8
null
https://www.nature.com/articles/s41586-023-06747-5
null
https://www.semanticscholar.org/paper/9b093787f9be3d480cd11f8ec6ca5b0e44050d6d
Automatically Solving Number Word Problems by Semantic Parsing and Reasoning
This paper presents a semantic parsing and reasoning approach to automatically solving math word problems. A new meaning representation language is designed to bridge natural language text and math expressions. A CFG parser is implemented based on 9,600 semi-automatically created grammar rules. We conduct experiments o...
A new meaning representation language is designed to bridge natural language text and math expressions and a CFG parser is implemented based on 9,600 semi-automatically created grammar rules.
# Automatically Solving Number Word Problems by Semantic Parsing and Reasoning **Shuming Shi[1], Yuehui Wang[2*], Chin-Yew Lin[1], Xiaojiang Liu[1] and Yong Rui[1]** 1 Microsoft Research ``` {shumings, cyl, xiaojl, yongrui}@microsoft.com ``` 2 University of Science and Technology of China ``` wyh9346...
[ "Shuming, Shi", "Yuehui, Wang", "Chin-Yew, Lin", "Xiaojiang, Liu", "Yong, Rui" ]
2015-01-01T00:00:00
null
false
154
19
null
http://aclweb.org/anthology/D15-1135
null
https://www.semanticscholar.org/paper/acfe603d8a90ee3d4187e426b26c4d7bf394b4c8
Mathematical discoveries from program search with large language models
Large Language Models (LLMs) have demonstrated tremendous capabilities in solving complex tasks, from quantitative reasoning to understanding natural language. However, LLMs sometimes suffer from confabulations (or hallucinations) which can result in them making plausible but incorrect statements [1,2]. This hinders th...
This work introduces FunSearch (short for searching in the function space), an evolutionary procedure based on pairing a pretrained LLM with a systematic evaluator that demonstrates the effectiveness of this approach to surpass the best-known results in important problems, pushing the boundary of existing LLM-based app...
null
[ "Alhussein, Fawzi", "Matej, Balog", "Omar, Fawzi", "Bernardino, Romera-Paredes", "M. Pawan, Kumar", "Mohammadamin, Barekatain", "Emilien, Dupont", "Alexander, Novikov", "Francisco J. R., Ruiz", "Jordan S., Ellenberg", "Pengming, Wang", "Pushmeet, Kohli" ]
2023-12-14T00:00:00
null
false
152
3
null
https://www.nature.com/articles/s41586-023-06924-6
null
https://www.semanticscholar.org/paper/d32ba88571141ed0ebe7aeefbaa4ccaf8cda7be3
Reasoning about Quantities in Natural Language
N/A
A computational approach is developed which is shown to successfully recognize and normalize textual expressions of quantities and is used to further develop algorithms to assist reasoning in the context of the aforementioned tasks.
null
[ "Subhro, Roy", "Tim, Vieira", "Dan, Roth" ]
2015-01-01T00:00:00
null
false
149
18
null
https://direct.mit.edu/tacl/article-abstract/doi/10.1162/tacl_a_00118/43260
null
https://www.semanticscholar.org/paper/187c25de08261760cf48b4e9dabf308d2f7f15d9
Solving Geometry Problems: Combining Text and Diagram Interpretation
N/A
GEOS is introduced, the first automated system to solve unaltered SAT geometry questions by combining text understanding and diagram interpretation, and it is shown that by integrating textual and visual information, GEOS boosts the accuracy of dependency and semantic parsing of the question text.
# Solving Geometry Problems: Combining Text and Diagram Interpretation **Minjoon Seo, Hannaneh Hajishirzi, Ali Farhadi, Oren Etzioni, Clint Malcolm** University of Washington, Allen Institute for Artificial Intelligence _{minjoon,hannaneh,clintm}@washington.edu,{alif,orene}@allenai.org_ **Abstract** This paper int...
[ "Minjoon, Seo", "Ali, Farhadi", "Hannaneh, Hajishirzi", "Oren, Etzioni", "Clint, Malcolm" ]
2015-09-01T00:00:00
null
false
149
12
null
https://aclanthology.org/D15-1171
null
https://www.semanticscholar.org/paper/c87dccf7c21e67679389f23f86f039cd96720c3f
Deep Network Guided Proof Search
N/A
Experimental evidence is given that with a hybrid, two-phase approach, deep learning based guidance can significantly reduce the average number of proof search steps while increasing the number of theorems proved.
null
[ "Cezary, Kaliszyk", "Sarah, Loos", "Geoffrey, Irving", "Christian, Szegedy" ]
2017-01-01T00:00:00
null
false
148
18
null
https://www.semanticscholar.org/paper/e77e00286a63a32dafb9629cd79f6a77bddb1941
null
https://www.semanticscholar.org/paper/e77e00286a63a32dafb9629cd79f6a77bddb1941
Translating a Math Word Problem to an Expression Tree
Sequence-to-sequence (SEQ2SEQ) models have been successfully applied to automatic math word problem solving. Despite its simplicity, a drawback still remains: a math word problem can be correctly solved by more than one equations. This non-deterministic transduction harms the performance of maximum likelihood estimatio...
By considering the uniqueness of expression tree, an equation normalization method to normalize the duplicated equations is proposed and it is shown that the ensemble model with equationnormalization significantly outperforms the previous state-of-the-art methods.
null
[ "Dongxiang, Zhang", "Lei, Wang", "Yan, Wang", "Deng, Cai", "Xiaojiang, Liu" ]
2018-11-14T00:00:00
null
false
144
20
null
https://www.semanticscholar.org/paper/Translating-a-Math-Word-Problem-to-an-Expression-Wang-Wang/6605bba6e0caabda06b090d67698a5683eba4dfa
null
https://www.semanticscholar.org/paper/6605bba6e0caabda06b090d67698a5683eba4dfa
Making Language Models Better Reasoners with Step-Aware Verifier
Few-shot learning is a challenging task that requires language models to generalize from limited examples. Large language models like GPT-3 and PaLM have made impressive progress in this area, but they still face difficulties in reasoning tasks such as GSM8K, a benchmark for arithmetic problems. To improve their reason...
This paper presents DiVeRSe (Diverse Verifier on Reasoning Step), a novel approach that further enhances the reasoning capability of language models and achieves new state-of-the-art results on six of eight reasoning benchmarks.
# Making Large Language Models Better Reasoners with Step-Aware Verifier **Yifei Li[1,2][∗], Zeqi Lin[2], Shizhuo Zhang[2], Qiang Fu[2], Bei Chen[2],** **Jian-Guang Lou[2], Weizhu Chen[2]** 1 National Key Laboratory for Multimedia Information Processing, School of Computer Science, Peking University 2 Microsoft Corpo...
[ "Yifei, Li", "Zeqi, Lin", "Shizhuo, Zhang", "Qiang, Fu", "Bei, Chen", "Jian-Guang, Lou", "Weizhu, Chen", "Anna, Rogers", "Jordan, Boyd-Graber", "Naoaki, Okazaki" ]
2023-07-01T00:00:00
ACL 2023 Long Papers
true
138
11
null
https://aclanthology.org/2023.acl-long.291
null
https://www.semanticscholar.org/paper/e826ac71dad8c4ce36d82fb7add43e3d306bb7e1
Reinforcement Learning of Theorem Proving
N/A
A theorem proving algorithm that uses practically no domain heuristics for guiding its connection-style proof search and solves within the same number of inferences over 40% more problems than a baseline prover, which is an unusually high improvement in this hard AI domain.
null
[ "Cezary, Kaliszyk", "Josef, Urban", "Henryk, Michalewski", "Miroslav, Olšák" ]
2018-01-01T00:00:00
NeurIPS 2018
true
135
13
[ "Mizar" ]
https://www.semanticscholar.org/paper/920febb03475b068286a855c10ea09b968fe7ee3
null
https://www.semanticscholar.org/paper/920febb03475b068286a855c10ea09b968fe7ee3
Template-Based Math Word Problem Solvers with Recursive Neural Networks
The design of automatic solvers to arithmetic math word problems has attracted considerable attention in recent years and a large number of datasets and methods have been published. Among them, Math23K is the largest data corpus that is very helpful to evaluate the generality and robustness of a proposed solution. The ...
This paper designs a recursive neural network to encode the quantity with Bi-LSTM and self attention, and infer the unknown operator nodes in a bottom-up manner, and establishes the superiority of the new framework as it improves the accuracy by a wide margin in two of the largest datasets.
null
[ "Jipeng, Zhang", "Dongxiang, Zhang", "Lei, Wang", "Lianli, Gao", "Xing, Xu", "Bing Tian, Dai", "Heng Tao, Shen" ]
2019-07-17T00:00:00
AAAI 2019 Natural Language Processing
false
134
22
null
https://ojs.aaai.org/index.php/AAAI/article/view/4697
null
https://www.semanticscholar.org/paper/9ff3afa6577d5469cd68c4991476f426efdf6bca
Rethinking with Retrieval: Faithful Large Language Model Inference
Despite the success of large language models (LLMs) in various natural language processing (NLP) tasks, the stored knowledge in these models may inevitably be incomplete, out-of-date, or incorrect. This motivates the need to utilize external knowledge to assist LLMs. Unfortunately, current methods for incorporating ext...
This work proposes a novel post-processing approach, rethinking with retrieval (RR), which retrieves relevant external knowledge based on the decomposed reasoning steps obtained from the chain-of-thought (CoT) prompting, which can produce more faithful explanations and improve the performance of LLMs.
## Rethinking with Retrieval: Faithful Large Language Model Inference **Hangfeng He[†][∗]** **Hongming Zhang[‡]** **Dan Roth[§]** _†University of Rochester_ _‡Tencent AI Lab, Seattle_ _§University of Pennsylvania_ hanfeng.he@rochester.edu, hongmzhang@global.tencent.com danroth@seas.upenn.edu **Abstract** Despite t...
[ "Hangfeng, He", "Hongming, Zhang", "Dan, Roth" ]
2022-12-31T00:00:00
null
false
131
2
null
http://arxiv.org/abs/2301.00303
https://arxiv.org/abs/2301.00303
https://www.semanticscholar.org/paper/490d8006851b1562cfd9ec1f057471f2868289d1
How well do Computers Solve Math Word Problems? Large-Scale Dataset Construction and Evaluation
No summary was provided.
A large-scale dataset is built which is more than 9 times the size of previous ones, and contains many more problem types, and semi-automatically obtained from community question-answering web pages.
# How Well Do Computers Solve Math Word Problems? Large-Scale Dataset Construction and Evaluation **Danqing Huang[1][∗], Shuming Shi[2], Chin-Yew Lin[2], Jian Yin[1]** **and Wei-Ying Ma[2]** 1 Sun Yat-sen University _{huangdq2@mail2,issjyin@mail2_ _}.sysu.edu.cn_ Microsoft Research _{shumings,cyl,wyma}@microsoft.com_...
[ "Danqing, Huang", "Shuming, Shi", "Chin-Yew, Lin", "Noah A., Smith", "Wei-Ying, Ma", "Katrin, Erk", "Jian, Yin" ]
2016-08-01T00:00:00
ACL 2016 Long Papers
true
130
21
null
https://aclanthology.org/P16-1084
null
https://www.semanticscholar.org/paper/09900b36ca35af12301507cf992675808a709838
The Unreliability of Explanations in Few-shot Prompting for Textual Reasoning
Does prompting a large language model (LLM) like GPT-3 with explanations improve in-context learning? We study this question on two NLP tasks that involve reasoning over text, namely question answering and natural language inference. We test the performance of four LLMs on three textual reasoning datasets using prompts...
This work studies two NLP tasks that involve reasoning over text, namely question answering and natural language inference, and shows that explanations judged by humans to be good--logically consistent with the input and the prediction--more likely cooccur with accurate predictions.
# The Unreliability of Explanations in Few-shot Prompting for Textual Reasoning **Xi Ye** **Greg Durrett** Department of Computer Science The University of Texas at Austin {xiye,gdurrett}@cs.utexas.edu **Abstract** Does prompting a large language model (LLM) like GPT-3 with explanations improve in-context learning? ...
[ "Xi, Ye", "Greg, Durrett" ]
2022-10-31T00:00:00
NeurIPS 2022 Poster
true
129
6
null
https://openreview.net/forum?id=Bct2f8fRd8S
https://arxiv.org/abs/2205.03401
https://www.semanticscholar.org/paper/9ffefdf1fcd780cb71450b0a7a29247c66aa87be
Inter-GPS: Interpretable Geometry Problem Solving with Formal Language and Symbolic Reasoning
Geometry problem solving has attracted much attention in the NLP community recently. The task is challenging as it requires abstract problem understanding and symbolic reasoning with axiomatic knowledge. However, current datasets are either small in scale or not publicly available. Thus, we construct a new large-scale ...
This work constructs a new largescale benchmark, Geometry3K, consisting of 3,002 geometry problems with dense annotation in formal language, and proposes a novel geometry solving approach with formal language and symbolic reasoning, called Interpretable Geometry Problem Solver (InterGPS).
## Inter-GPS: Interpretable Geometry Problem Solving with Formal Language and Symbolic Reasoning **Pan Lu[1][∗], Ran Gong[1][∗], Shibiao Jiang[2][∗], Liang Qiu[1], Siyuan Huang[1],** **Xiaodan Liang[3], Song-Chun Zhu[1][,][4][,][5][,][6]** 1Center for Vision, Cognition, Learning and Autonomy, UCLA 2College of Comput...
[ "Pan, Lu", "Siyuan, Huang", "Ran, Gong", "Shibiao, Jiang", "Liang, Qiu", "Song-Chun, Zhu", "Xiaodan, Liang" ]
2021-07-20T00:00:00
ACL 2021 Long Papers
true
126
12
null
http://arxiv.org/abs/2105.04165
https://arxiv.org/abs/2105.04165
https://www.semanticscholar.org/paper/fb1c90806fc5ec72987f58110aa255edbce6620d
Can Neural Networks Understand Logical Entailment?
N/A
Results show that convolutional networks present the wrong inductive bias for this class of problems relative to LSTM RNNs, tree-structured neural networks outperform LSTS due to their enhanced ability to exploit the syntax of logic, and PossibleWorldNets outperform all benchmarks.
null
[ "David, Saxton", "Richard, Evans", "David, Amos", "Edward, Grefenstette", "Pushmeet, Kohli" ]
2018-01-01T00:00:00
null
false
125
6
null
https://www.semanticscholar.org/paper/bdea8b6ceabaeb86bd23c2d2585da1ff3858d968
null
https://www.semanticscholar.org/paper/bdea8b6ceabaeb86bd23c2d2585da1ff3858d968
Premise Selection for Theorem Proving by Deep Graph Embedding
We propose a deep learning-based approach to the problem of premise selection: selecting mathematical statements relevant for proving a given conjecture. We represent a higher-order logic formula as a graph that is invariant to variable renaming but still fully preserves syntactic and semantic information. We then embe...
A deep learning-based approach to the problem of premise selection: selecting mathematical statements relevant for proving a given conjecture by representing a higher-order logic formula as a graph that is invariant to variable renaming but still fully preserves syntactic and semantic information.
## Premise Selection for Theorem Proving by Deep Graph Embedding **Mingzhe Wang[∗]** **Yihe Tang[∗]** **Jian Wang** **Jia Deng** University of Michigan, Ann Arbor **Abstract** We propose a deep learning-based approach to the problem of premise selection: selecting mathematical statements relevant for proving a give...
[ "Mingzhe, Wang", "Jia, Deng", "Yihe, Tang", "Jian, Wang" ]
2017-09-28T00:00:00
NeurIPS 2017
true
124
12
null
http://arxiv.org/abs/1709.09994
null
https://www.semanticscholar.org/paper/15bce8864a15c20cf483e18099f3601b94673565
A Neural Network Solves, Explains, and Generates University Math Problems by Program Synthesis and Few-Shot Learning at Human Level
We demonstrate that a neural network pre-trained on text and fine-tuned on code solves mathematics course problems, explains solutions, and generates new questions at a human level. We automatically synthesize programs using few-shot learning and OpenAI's Codex transformer and execute them to solve course problems at 8...
This work solves university-level mathematics courses and improves upon state-of-the-art, increasing automatic accuracy on randomly sampled questions on a benchmark by order of magnitude.
# A neural network solves, explains, and generates university math problems by program synthesis and few-shot learning at human level Iddo Drori[a,b,1], Sarah Zhang[c], Reece Shuttleworth[a], Leonard Tang[d], Albert Lu[a], Elizabeth Ke[a], Kevin Liu[a], Linda Chen[a], Sunny Tran[a], Newman Cheng[b], Roman Wang[b], Nik...
[ "Iddo, Drori", "Sarah, Zhang", "Reece, Shuttleworth", "Leonard, Tang", "Albert, Lu", "Elizabeth, Ke", "Kevin, Liu", "Linda, Chen", "Sunny, Tran", "Newman, Cheng" ]
2022-01-01T00:00:00
null
false
123
8
null
https://arxiv.org/abs/2112.15594
https://arxiv.org/abs/2112.15594
https://www.semanticscholar.org/paper/257aee73d83a87921fd2d56b524de394dcf6a264
DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
Mathematical reasoning poses a significant challenge for language models due to its complex and structured nature. In this paper, we introduce DeepSeekMath 7B, which continues pre-training DeepSeek-Coder-Base-v1.5 7B with 120B math-related tokens sourced from Common Crawl, together with natural language and code data. ...
Group Relative Policy Optimization (GRPO), a variant of Proximal Policy Optimization (PPO), that enhances mathematical reasoning abilities while concurrently optimizing the memory usage of PPO is introduced.
## DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models Zhihong Shao[1,2][∗†], Peiyi Wang[1,3][∗†], Qihao Zhu[1,3][∗†], Runxin Xu[1], Junxiao Song[1] Mingchuan Zhang[1], Y.K. Li[1], Y. Wu[1], Daya Guo[1][∗] 1DeepSeek-AI, 2Tsinghua University, 3Peking University ``` {zhihongshao,wangpei...
[ "Zhihong, Shao", "Peiyi, Wang", "Qihao, Zhu", "Runxin, Xu", "Junxiao, Song", "Mingchuan, Zhang", "Y. K., Li", "Y., Wu", "Daya, Guo" ]
2024-02-06T00:00:00
null
false
123
8
[ "Isabelle" ]
http://arxiv.org/abs/2402.03300
https://arxiv.org/abs/2402.03300
https://www.semanticscholar.org/paper/35b142ea69598e6241f0011312128031df55895c
ToolkenGPT: Augmenting Frozen Language Models with Massive Tools via Tool Embeddings
Integrating large language models (LLMs) with various tools has led to increased attention in the field. Existing approaches either involve fine-tuning the LLM, which is both computationally costly and limited to a fixed set of tools, or prompting LLMs by in-context tool demonstrations. Although the latter method offer...
The proposed ToolkenGPT offers the flexibility to plug in an arbitrary number of tools by expanding the set of toolkens on the fly and improves tool use by allowing extensive demonstration data for learning the toolken embeddings.
## ToolkenGPT: Augmenting Frozen Language Models with Massive Tools via Tool Embeddings **Shibo Hao[1], Tianyang Liu[1], Zhen Wang[1][,][ 2], Zhiting Hu[1]** 1UC San Diego, 2Mohamed bin Zayed University of Artificial Intelligence ``` {s5hao, til040, zhw085, zhh019}@ucsd.edu ``` **Abstract** Augmenting lar...
[ "Shibo, Hao", "Zhen, Wang", "Tianyang, Liu", "Zhiting, Hu" ]
2024-01-15T00:00:00
NeurIPS 2023
true
123
2
null
http://arxiv.org/abs/2305.11554
https://arxiv.org/abs/2305.11554
https://www.semanticscholar.org/paper/c7a3f9cc61cfafdc307f8ae24430b6b1121f9b2c
Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks
The impressive performance of recent language models across a wide range of tasks suggests that they possess a degree of abstract reasoning skills. Are these skills general and transferable, or specialized to specific tasks seen during pretraining? To disentangle these effects, we propose an evaluation framework based ...
An evaluation framework based on “counterfactual” task variants that deviate from the default assumptions underlying standard tasks suggests that while current LMs may possess abstract task-solving skills to an extent, they often also rely on narrow, non-transferable procedures for task-solving.
# Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks **Zhaofeng Wu[ã]** **Linlu Qiu[ã]** **Alexis Ross[ã]** **Ekin Akyürek[ã]** **Boyuan Chen[ã]** **Bailin Wang[ã]** **Najoung Kim[æ]** **Jacob Andreas[ã]** **Yoon Kim[ã]** ãMIT æBoston University zfw@csai...
[ "Boyuan, Chen", "Ekin, Akyürek", "Bailin, Wang", "Zhaofeng, Wu", "Yoon, Kim", "Najoung, Kim", "Linlu, Qiu", "Alexis, Ross", "Kevin, Duh", "Jacob, Andreas", "Helena, Gomez", "Steven, Bethard" ]
2024-06-01T00:00:00
NAACL 2024 Long Papers
true
120
2
null
https://aclanthology.org/2024.naacl-long.102
https://arxiv.org/abs/2307.02477
https://www.semanticscholar.org/paper/f8e99be4f9a01761fab74bade2c3c18de9fc686b
Learning to Prove Theorems via Interacting with Proof Assistants
N/A
ASTactic, a deep learning-based model that generates tactics as programs in the form of abstract syntax trees (ASTs) can generate effective tactics and can be used to prove new theorems not previously provable by automated methods.
null
[ "Kaiyu, Yang", "Jia, Deng" ]
2019-01-01T00:00:00
ICML 2019
true
117
22
[ "Coq" ]
https://www.semanticscholar.org/paper/a596f03145285cd05a6ca57a4e25418b23b24976
null
https://www.semanticscholar.org/paper/a596f03145285cd05a6ca57a4e25418b23b24976
Large Language Model Guided Tree-of-Thought
In this paper, we introduce the Tree-of-Thought (ToT) framework, a novel approach aimed at improving the problem-solving capabilities of auto-regressive large language models (LLMs). The ToT technique is inspired by the human mind's approach for solving complex reasoning tasks through trial and error. In this process, ...
Experimental results show that the ToT framework can significantly increase the success rate of Sudoku puzzle solving and the implemented ToT-based Sudoku solver is available on GitHub.
## Large Language Model Guided Tree-of-Thought **Jieyi Long** Theta Labs, Inc. San Jose, CA 95128 ``` jieyi@thetalabs.org ``` **Abstract** In this paper, we introduce the Tree-of-Thought (ToT) framework, a novel approach aimed at improving the problem-solving capabilities of auto-regressive large langu...
[ "Jieyi, Long" ]
2023-05-14T00:00:00
null
false
116
5
null
http://arxiv.org/abs/2305.08291
https://arxiv.org/abs/2305.08291
https://www.semanticscholar.org/paper/bda605928d6ebe4db906e69ab5d343df75918727
TALM: Tool Augmented Language Models
Transformer based language models (LMs) demonstrate increasing performance with scale across a wide variety of tasks. Scale alone however cannot enable models to solve tasks that require access to ephemeral, changing, or private data that was unavailable at training time. Many useful tasks may also benefit from LMs bei...
TALM is presented, combining a text-only approach to augment language models with non-differentiable tools, and an iterative"self-play"technique to bootstrap performance starting from few tool demonstrations, suggesting that Tool Augmented Language Models are a promising direction to enrich LMs' capabilities, with less...
## TALM: Tool Augmented Language Models **Aaron Parisi** **Yao Zhao** **Noah Fiedel** _{aarontp,yaozhaoyz,nfiedel}_ @google.com **Abstract** Transformer based language models (LMs) demonstrate increasing performance with scale across a wide variety of tasks. Scale alone however cannot enable models to solve tasks...
[ "Aaron, Parisi", "Yao, Zhao", "Noah, Fiedel" ]
2022-05-24T00:00:00
null
false
115
7
null
http://arxiv.org/abs/2205.12255
https://arxiv.org/abs/2205.12255
https://www.semanticscholar.org/paper/354bf043179e3e9f05df73e3f04517e53c326d1f
NumNet: Machine Reading Comprehension with Numerical Reasoning
Numerical reasoning, such as addition, subtraction, sorting and counting is a critical skill in human’s reading comprehension, which has not been well considered in existing machine reading comprehension (MRC) systems. To address this issue, we propose a numerical MRC model named as NumNet, which utilizes a numerically...
A numerical MRC model named as NumNet is proposed, which utilizes a numerically-aware graph neural network to consider the comparing information and performs numerical reasoning over numbers in the question and passage, outperforming all existing machine reading comprehension models by considering the numerical relatio...
# NumNet: Machine Reading Comprehension with Numerical Reasoning **Qiu Ran[1][∗], Yankai Lin[1][∗], Peng Li[1], Jie Zhou[1], Zhiyuan Liu[2]** 1Pattern Recognition Center, WeChat AI, Tencent Inc, China 2Department of Computer Science and Technology, Tsinghua University, Beijing, China Institute for Artificial Intellig...
[ "Qiu, Ran", "Jie, Zhou", "Peng, Li", "Zhiyuan, Liu", "Yankai, Lin" ]
2019-01-01T00:00:00
EMNLP 2019 Main
true
112
8
null
https://www.aclweb.org/anthology/D19-1251
null
https://www.semanticscholar.org/paper/730043364aed106241ef18ab3e3b5e316802a254
LeanDojo: Theorem Proving with Retrieval-Augmented Language Models
Large language models (LLMs) have shown promise in proving formal theorems using proof assistants such as Lean. However, existing methods are difficult to reproduce or build on, due to private code, data, and large compute requirements. This has created substantial barriers to research on machine learning methods for t...
This paper introduces LeanDojo: an open-source Lean playground consisting of toolkits, data, models, and benchmarks, and develops ReProver (Retrieval-Augmented Prover): an LLM-based prover augmented with retrieval for selecting premises from a vast math library.
## LeanDojo: Theorem Proving with Retrieval-Augmented Language Models **Kaiyu Yang[1], Aidan M. Swope[2], Alex Gu[3], Rahul Chalamala[1], Peiyang Song[4],** **Shixing Yu[5], Saad Godil[∗], Ryan Prenger[2], Anima Anandkumar[1][,][2]** 1Caltech, 2NVIDIA, 3MIT, 4 UC Santa Barbara, 5UT Austin ``` https://lea...
[ "Kaiyu, Yang", "Aidan M., Swope", "Alex, Gu", "Rahul, Chalamala", "Peiyang, Song", "Shixing, Yu", "Saad, Godil", "Ryan, Prenger", "Anima, Anandkumar" ]
2023-06-27T00:00:00
NeurIPS 2023
true
111
12
[ "Lean" ]
http://arxiv.org/abs/2306.15626
https://arxiv.org/abs/2306.15626
https://www.semanticscholar.org/paper/87875a07976c26f82705de1fc70041169e5d652b
The Gap of Semantic Parsing: A Survey on Automatic Math Word Problem Solvers
Solving mathematical word problems (MWPs) automatically is challenging, primarily due to the semantic gap between human-readable words and machine-understandable logics. Despite the long history dated back to the 1960s, MWPs have regained intensive attention in the past few years with the advancement of Artificial Inte...
This survey focuses on algebraic word problems, summarize their extracted features and proposed techniques to bridge the semantic gap, and compare their performance in the publicly accessible datasets.
null
[ "Dongxiang, Zhang", "Lei, Wang", "Luming, Zhang", "Bing Tian, Dai", "Heng Tao, Shen" ]
2020-09-01T00:00:00
null
false
110
7
null
https://ieeexplore.ieee.org/document/8703135/
null
https://www.semanticscholar.org/paper/a810362ca5284e5894b07abb76b8f647802fe32c
Large Language Models are Better Reasoners with Self-Verification
Recently, with the chain of thought (CoT) prompting, large language models (LLMs), e.g., GPT-3, have shown strong reasoning ability in several natural language processing tasks such as arithmetic, commonsense, and logical reasoning. However, LLMs with CoT require multi-step prompting and multi-token prediction, which i...
This paper proposes and proves that LLMs also have similar self-verification abilities, and takes the conclusion obtained by CoT as one of the conditions for solving the original problem.
## Large Language Models are Better Reasoners with Self-Verification **Yixuan Weng[1][∗], Minjun Zhu[1][,][2][∗], Fei Xia[1][,][2], Bin Li[3],** **Shizhu He[1][,][2][†], Shengping Liu[4], Bin Sun[3], Kang Liu[1][,][2][,][5], Jun Zhao[1][,][2]** 1 The Laboratory of Cognition and Decision Intelligence for Complex Syste...
[ "Jun, Zhao", "Yixuan, Weng", "Minjun, Zhu", "Bin, Li", "Shizhu, He", "Fei, Xia", "Kang, Liu", "Bin, Sun", "Shengping, Liu" ]
2022-01-01T00:00:00
EMNLP 2023 Findings
false
109
5
null
https://arxiv.org/abs/2212.09561
https://arxiv.org/abs/2212.09561
https://www.semanticscholar.org/paper/7715ba5e75f5256e1061c7473afe61bb0dbb9065
Autoformalization with Large Language Models
Autoformalization is the process of automatically translating from natural language mathematics to formal specifications and proofs. A successful autoformalization system could advance the fields of formal verification, program synthesis, and artificial intelligence.While the long-term goal of autoformalization seemed ...
It is shown large language models provide new prospects towards the long-term goal of autoformalization, and the surprising observation that LLMs can correctly translate a significant portion of mathematical competition problems perfectly to formal specifications in Isabelle/HOL.
### Autoformalization with Large Language Models Yuhuai Wu[1][,][2][†] Albert Q. Jiang[3] Wenda Li[3] Markus N. Rabe[1] Charles Staats[1] Mateja Jamnik[3] Christian Szegedy[1] 1Google Research 2Stanford University 3University of Cambridge Abstract Autoformalization is the process of automatically translating fro...
[ "Albert Q., Jiang", "Wenda, Li", "Yuhuai, Wu", "Markus N., Rabe", "Christian, Szegedy", "Charles, Staats", "Mateja, Jamnik" ]
2022-05-01T00:00:00
NeurIPS 2022
true
108
16
[ "Lean" ]
https://arxiv.org/abs/2205.12615
https://arxiv.org/abs/2205.12615
https://www.semanticscholar.org/paper/c28e95a06dfcf13fc65a1cac83722f53e34f12a5
Distilling Reasoning Capabilities into Smaller Language Models
Step-by-step reasoning approaches like chain of thought (CoT) have proved to be very effective in inducing reasoning capabilities in large language models. However, the success of the CoT approach is fundamentally tied to the model size, and billion parameter-scale models are often needed to get CoT to work. In this pa...
This work proposes an alternative reasoning scheme, Socratic CoT, that learns a decomposition of the original problem into a sequence of subproblems and uses it to guide the intermediate reasoning steps, and uses this to train a combination of two small distilled models.
**Distilling Reasoning Capabilities into Smaller Language Models** **Kumar Shridhar[∗]** **Alessandro Stolfo[∗]** **Mrinmaya Sachan** Department of Computer Science, ETH Z¨urich _{shkumar, stolfoa}@ethz.ch_ Step-by-step reasoning approaches like chain that much white fiber. How many bolts in of thought (CoT) have pro...
[ "Alessandro, Stolfo", "Kumar, Shridhar", "Mrinmaya, Sachan", "Anna, Rogers", "Jordan, Boyd-Graber", "Naoaki, Okazaki" ]
2023-07-01T00:00:00
ACL 2023 Findings
false
108
6
null
https://aclanthology.org/2023.findings-acl.441
https://arxiv.org/abs/2212.00193
https://www.semanticscholar.org/paper/8fd462f6248d5e3f1b6602697c09489086b5655f
Explanations from Large Language Models Make Small Reasoners Better
Integrating free-text explanations to in-context learning of large language models (LLM) is shown to elicit strong reasoning capabilities along with reasonable explanations. In this paper, we consider the problem of leveraging the explanations generated by LLM to improve the training of small reasoners, which are more ...
This paper considers the problem of leveraging the explanations generated by LLM to improve the training of small reasoners, which are more favorable in real-production deployment due to their low cost.
### Explanations from Large Language Models Make Small Reasoners Better **Shiyang Li[1], Jianshu Chen[2], Yelong Shen[3], Zhiyu Chen[1], Xinlu Zhang[1]** **, Zekun Li[1]** **Hong Wang[1], Jing Qian[1], Baolin Peng[3], Yi Mao[3], Wenhu Chen[4]** and Xifeng Yan[1] 1University of California, Santa Barbara 2Tencent AI L...
[ "Zhiyu, Chen", "Jing, Qian", "Baolin, Peng", "Shiyang, Li", "Hong, Wang", "Jianshu, Chen", "Zekun, Li", "Yelong, Shen", "Wenhu, Chen", "Xifeng, Yan", "Xinlu, Zhang", "Yi, Mao" ]
2022-10-13T00:00:00
null
false
108
6
null
http://arxiv.org/abs/2210.06726
https://arxiv.org/abs/2210.06726
https://www.semanticscholar.org/paper/7d29a84a589aa5655e5d3fed8d725ea472816599
REFINER: Reasoning Feedback on Intermediate Representations
Language models (LMs) have recently shown remarkable performance on reasoning tasks by explicitly generating intermediate inferences, e.g., chain-of-thought prompting. However, these intermediate inference steps may be inappropriate deductions from the initial context and lead to incorrect final predictions. Here we in...
REFINER is a framework for finetuning LMs to explicitly generate intermediate reasoning steps while interacting with a critic model that provides automated feedback on the reasoning that provides structured feedback that the reasoning LM uses to iteratively improve its intermediate arguments.
### REFINER: Reasoning Feedback on Intermediate Representations **Debjit Paul[♠], Mete Ismayilzada[♠], Maxime Peyrard[♢∗], Beatriz Borges[♠],** **Antoine Bosselut[♠], Robert West[♠], Boi Faltings[♠]** _♠EPFL_ _♢Université Grenoble Alpes, CNRS, Grenoble INP, LIG_ {firstname.lastname}@epfl.ch **Abstract** Language m...
[ "Debjit, Paul", "Mete, Ismayilzada", "Maxime, Peyrard", "Beatriz, Borges", "Antoine, Bosselut", "Robert, West", "Boi, Faltings" ]
2024-02-04T00:00:00
null
false
108
17
null
http://arxiv.org/abs/2304.01904
https://arxiv.org/abs/2304.01904
https://www.semanticscholar.org/paper/c715914c388fa64dd8686cd8755e5adfebbf2388