Ring-lite-rl-data: Reinforcement Learning Dataset for Math and Code

47,400 curated problems spanning mathematics and programming domains specifically formatted for reinforcement learning. The collection includes 39,000 math problems from sources like AoPS and DeepMath-103K, alongside approximately 8,400 coding challenges.

Use Cases

• Train reinforcement learning agents for mathematical reasoning using the 39,000 curated math problems
• Fine-tune large language models on competitive programming tasks using the 8,400 code-domain entries
• Benchmark model performance on contest-level mathematics using the AoPS-sourced problem sets

Strengths

• 39,000+ mathematics problems sourced from BigMath, DeepScaleR, DAPO, and DeepMath-103K
• 8,400 programming problems curated for code-specific reinforcement learning
• Includes contest-level math problems from the Art of Problem Solving (AoPS) collections