{"data":{"kind":"file","path":"README.md","version_id":"yp2i5dqxha1uxsoifzh0dl2r","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1628,"modified_at":"2026-08-09T07:24:50.262000","content_hash":"3cf0c7b17fa995cc72d9d1e4d787a05396edbdc9bb9823487738800b8717daa1"},"entries":[],"content":"# Mastermind / Code-Breaking Game\n\nA logic and deduction environment where the model must crack secret codes in the classic Mastermind board game.\n\n## Game Rules\n\n- A secret code consists of 4 colored pegs chosen from 6 colors: Red (R), Blue (B), Green (G), Yellow (Y), Orange (O), Purple (P)\n- Colors may repeat in the code\n- After each guess, feedback is given:\n  - **Black pegs (B)**: correct color in the correct position\n  - **White pegs (W)**: correct color but wrong position\n- The goal is to deduce the secret code from a set of guess-feedback clues\n\n## Task Format\n\nEach puzzle presents:\n1. The game setup (4 pegs, 6 colors)\n2. A series of guesses with their feedback\n3. The model must output the secret code as 4 space-separated color letters\n\n## Difficulty Levels\n\n- **Easy (level 0)**: 3-4 clues, straightforward elimination\n- **Medium (level 1)**: 2 clues, requires process-of-elimination reasoning\n- **Hard (level 2)**: 1-2 clues with ambiguous information, deep logical inference needed\n\n## Scoring\n\n- **Code Correctness** (weight 1.0): Exact match of all 4 positions\n- **Partial Credit** (weight 0.3): Fraction of correctly identified positions\n- **Reasoning Quality** (weight 0.5): Model demonstrates systematic deduction steps\n\n## Example\n\n```\nClues:\n  Guess: R B G Y -> 1 black, 1 white\n  Guess: R Y O B -> 0 black, 2 white\n  Guess: G O Y R -> 1 black, 1 white\n\nWhat is the secret code?\nAnswer: B R Y G\n```\n\n## Environment Design\n\nThis environment tests:\n- Logical deduction and constraint satisfaction\n- Process of elimination\n- Systematic reasoning under uncertainty\n- Color/position constraint propagation\n","encoding":"utf-8","truncated":false,"total_bytes":1628},"status":null}