Update README.md

README.md

@@ -4,7 +4,7 @@ language:
 - en
 pretty_name: MeetingToM
 size_categories:
-- n<1K
+- 1K<n<10K
 tags:
 - multimodal
 - theory-of-mind
@@ -35,17 +35,19 @@ configs:
 [GitHub](https://github.com/oliviaziyi/MeetingToM) ·
 [Project Page](https://oliviaziyi.github.io/MeetingToM-Project-Page/)
 
-MeetingToM is a multimodal benchmark for evaluating **Theory-of-Mind (ToM) reasoning in multi-party meetings**. It studies social reasoning at three levels: individual mental states, interpersonal relations, and group-level consensus.
+MeetingToM is a multimodal benchmark for evaluating **Theory-of-Mind (ToM) reasoning in multi-party meetings**. It studies social reasoning at three complementary levels: individual mental states, interpersonal relations, and group-level consensus.
 
-The benchmark contains **900 instances** and **1,800 gold answers** across three tasks.
+The released benchmark contains **900 source bundles**, expanded into **1,200 evaluation records** with **1,800 gold answers**.
 
-| Task | Level | Q1 | Q2 | Instances |
-|---|---|---|---|---:|
-| **STATE** | Individual | Early mental state | Late mental state | 300 |
-| **YOU** | Interpersonal | Addressee | Conversational stance | 300 |
-| **CONSENSUS** | Group | Consensus quality | Dissenter | 300 |
+| Task | Level | Source bundles | Evaluation records | Questions | Gold answers |
+|---|---|---:|---:|---|---:|
+| **STATE** | Individual | 300 | 600 | Q1: Mental state | 600 |
+| **YOU** | Interpersonal | 300 | 300 | Q1: Addressee; Q2: Conversational stance | 600 |
+| **CONSENSUS** | Group | 300 | 300 | Q1: Consensus quality; Q2: Dissenter | 600 |
 
-> **Note on media.** MeetingToM is constructed from the AMI Meeting Corpus. AMI-derived video and audio are not redistributed in this repository. Users should obtain authorized AMI access separately and reconstruct benchmark media locally using the released metadata and scripts.
+For STATE, each source bundle contributes two independently evaluated **5-second video clips**, resulting in 600 evaluation records. The same mental-state question is asked for every STATE record.
+
+> **Note on media.** MeetingToM is constructed from the AMI Meeting Corpus. AMI-derived video and audio are not redistributed in this repository. Users should obtain authorized access to AMI separately and reconstruct benchmark media locally using the released metadata and scripts.
 
 ---
 
@@ -59,7 +61,15 @@ you
 consensus
 ```
 
-Each configuration contains one `test` split with 300 instances.
+Each configuration contains one `test` split:
+
+```text
+state        600 evaluation records
+you          300 evaluation records
+consensus    300 evaluation records
+```
+
+The configurations can be loaded separately with 🤗 Datasets:
 
 ```python
 from datasets import load_dataset
@@ -81,23 +91,28 @@ state = load_dataset("OliviaWang1101/MeetingToM")
 
 ### 👤 STATE
 
-STATE evaluates how a target participant's mental state changes within a short meeting segment.
+STATE evaluates whether a model can infer the mental state of a target participant from a short meeting clip.
+
+Each STATE evaluation record contains an independently evaluated **5-second video clip** of the target participant together with the corresponding meeting audio.
+
+The model answers the same mental-state question for every clip:
 
-- **Q1:** mental state during the first 5 seconds of the source window.
-- **Q2:** mental state during the last 5 seconds of the source window.
-- **Media:** target participant's close-up view with the corresponding meeting audio.
+- **Q1 — Mental state:** infer the target participant's current cognitive or attentional state.
+
+Each source bundle contributes two independent STATE evaluation records, producing **600 STATE records** in total.
 
 Example:
 
 ```json
 {
-  "id": "state:ES2002a_state0004",
+  "id": "state:ES2002a_state0004_w0",
   "task": "state",
-  "bundle_name": "ES2002a_state0004",
+  "bundle_name": "ES2002a_state0004_w0",
+  "source_bundle_name": "ES2002a_state0004",
   "session_id": "ES2002a",
+  "window_index": 0,
   "answers": {
-    "Q1": "COGNITIVE_CONFLICT",
-    "Q2": "FOCUSED_LISTENING"
+    "Q1": "COGNITIVE_CONFLICT"
   }
 }
 ```
@@ -160,16 +175,26 @@ Example:
 
 ## Data format
 
-All three annotation files share the same basic schema:
+The three annotation files share several common fields:
 
 | Field | Description |
 |---|---|
-| `id` | Unique benchmark identifier |
+| `id` | Unique evaluation-record identifier |
 | `task` | `state`, `you`, or `consensus` |
-| `bundle_name` | Benchmark bundle identifier |
+| `bundle_name` | Released evaluation-record identifier |
 | `session_id` | Source AMI meeting session |
-| `answers.Q1` | Gold answer for Q1 |
-| `answers.Q2` | Gold answer for Q2 |
+| `answers` | Gold answer dictionary |
+
+STATE records additionally contain:
+
+| Field | Description |
+|---|---|
+| `source_bundle_name` | Original STATE source-bundle identifier |
+| `window_index` | Index distinguishing the two STATE records derived from the same source bundle |
+
+For STATE, `answers` contains a single `Q1` mental-state label.
+
+For YOU and CONSENSUS, `answers` contains both `Q1` and `Q2`.
 
 The Hugging Face repository is organized as:
 
@@ -190,18 +215,22 @@ MeetingToM/
 
 ## 🎬 Reconstruction metadata
 
-`metadata/reconstruction.jsonl` contains one reconstruction specification for each of the 900 benchmark instances.
+`metadata/reconstruction.jsonl` contains **1,200 reconstruction specifications**, aligned one-to-one with the 1,200 released evaluation records.
+
+The metadata provides the information needed to reconstruct benchmark media from an authorized local copy of the AMI Meeting Corpus.
+
+Depending on the task, reconstruction metadata includes information such as:
 
-Depending on the task, it records:
+- AMI session and source timestamps;
+- required camera views;
+- target participant and view information;
+- source-window specifications;
+- 2×2 mosaic layout;
+- session-specific mappings between close-up views and participant identities.
 
-- AMI session and source timestamps,
-- required camera views,
-- target participant information,
-- early and late STATE windows,
-- 2×2 mosaic layout,
-- session-specific mapping between close-up views and participant identities.
+For STATE, each reconstruction entry corresponds to one independent **5-second video clip**.
 
-For YOU-Q2 and CONSENSUS, the mosaic layout is:
+For YOU-Q2 and CONSENSUS, the close-up mosaic uses the following layout:
 
 ```text
 Closeup1 | Closeup2
@@ -215,7 +244,7 @@ The reconstruction metadata does **not** contain gold answers.
 
 The reconstruction code is maintained in the [GitHub repository](https://github.com/oliviaziyi/MeetingToM).
 
-After obtaining authorized AMI media, a single instance can be reconstructed with:
+After obtaining authorized AMI media, a single STATE evaluation record can be reconstructed with:
 
 ```bash
 python scripts/reconstruct.py \
@@ -223,10 +252,10 @@ python scripts/reconstruct.py \
   --audio_root /path/to/HeadsetAudio \
   --metadata metadata/reconstruction.jsonl \
   --output_dir reconstructed \
-  --id state:ES2002a_state0004
+  --id state:ES2002a_state0004_w0
 ```
 
-To reconstruct all instances:
+To reconstruct all released evaluation records:
 
 ```bash
 python scripts/reconstruct.py \
@@ -243,7 +272,7 @@ Reconstructed AMI-derived media should remain local and should not be redistribu
 
 ## Evaluation
 
-The official evaluator is available in the GitHub repository under:
+The official evaluator is maintained in the GitHub repository under:
 
 ```text
 evaluation/evaluate.py
@@ -260,7 +289,9 @@ The benchmark reports the following core metrics:
 | CONSENSUS-Q2 | Conditional Accuracy |
 | CONSENSUS | Two-step Points Accuracy |
 
-For CONSENSUS, Q2 is evaluated on instances where Q1 is correct. The point-based score is:
+STATE metrics are computed over the **600 independent STATE evaluation records**.
+
+For CONSENSUS, Q2 is evaluated on records where Q1 is predicted correctly. The point-based score is:
 
 ```text
 Q1 incorrect                     -> 0 points
@@ -272,14 +303,28 @@ See the GitHub repository for the complete evaluation protocol and prediction fo
 
 ---
 
+## Dataset integrity
+
+The public release contains:
+
+```text
+Source bundles          900
+Evaluation records    1,200
+Gold answers          1,800
+```
+
+The 1,200 released evaluation-record IDs align exactly with the 1,200 reconstruction metadata entries.
+
+---
+
 ## Source media
 
 MeetingToM is based on the **AMI Meeting Corpus**.
 
-This repository releases:
+This Hugging Face repository releases:
 
-- benchmark annotations,
-- reconstruction metadata,
+- benchmark annotations;
+- reconstruction metadata;
 - reconstruction summary.
 
 It does not release AMI video/audio or reconstructed clips and mosaics.