Skip to main content

Mathematics

Critic-Discernment Game

Critic-Discernment Game (CDG): a self-play reinforcement-learning training strategy that improves the rationality of an LLM reasoning process. Released lighteval detail files give per-(model, question) binary correctness (extractive_match) for the baseline Llama-3.1-8B-Instruct and the CDG-trained prover on GSM8K and MATH500.

1,819items
2subjects
100%observed
unknownlicense
mathematicsdomain
textmodality

Response matrix

Fit to width. Hover for subject & item; click a cell for details.

Critic-Discernment Game response matrix: AI models (rows) against items (columns)
Correct (1)Incorrect (0)Unobserved

Scale: 1 = correct · 0 = incorrect

Subjects

  1. 1Llama-3.1-8B-Instruct-CDG0.691
  2. 2Llama-3.1-8B-Instruct0.6811