Update README.md
Browse files
README.md
CHANGED
|
@@ -16,6 +16,7 @@ ArithMark 2.0 is a procedurally generated benchmark for evaluating integer arith
|
|
| 16 |
The benchmark is designed for base-model log-likelihood scoring. It does not require instruction following, chain-of-thought, or generated explanations. Random chance is 25%.
|
| 17 |
|
| 18 |
The included script benchmark_arithmark-2.0.py can be used to run the benchmark.
|
|
|
|
| 19 |
---
|
| 20 |
|
| 21 |
## Baseline Results
|
|
|
|
| 16 |
The benchmark is designed for base-model log-likelihood scoring. It does not require instruction following, chain-of-thought, or generated explanations. Random chance is 25%.
|
| 17 |
|
| 18 |
The included script benchmark_arithmark-2.0.py can be used to run the benchmark.
|
| 19 |
+
|
| 20 |
---
|
| 21 |
|
| 22 |
## Baseline Results
|