Pinned Loading

1. tendem-evaluation tendem-evaluation Public

   Tendem hybrid AI+Human system benchmarking

   Python 2

2. beemo beemo Public

   Benchmark for fine-grained machine-generated text detection. 6.5k texts written by humans, generated by ten open-source instruction-finetuned LLMs and edited by expert annotators.

   9 2

3. u-math u-math Public

   Official evaluation code for the U-MATH and μ-MATH benchmarks. These datasets are designed to test the mathematical reasoning and meta-evaluation capabilities of LLMs on university-level problems.

   Python 11 3

4. crowd-kit crowd-kit Public

   Control the quality of your labeled data with the Python tools you already know.

   Python 243 21