Code

The 'Code' dataset/benchmark contains programming-related tasks and is used to evaluate the reasoning capabilities of Large Language Models (LLMs) in generating and understanding code.

Papers using Code (3)

ResRL: Boosting LLM Reasoning via Negative Sample Projection Residual Reinforcement Learning2026Unforgotten Safety: Preserving Safety Alignment of Large Language Models with Continual Learning2025LongAttn: Selecting Long-context Training Data via Token-level Attention2025