HumanEval

The 'HumanEval' dataset is a benchmark that contains programming problems used to evaluate the performance of large reasoning models in generating code solutions.

Papers using HumanEval (3)

DistillGuard: Evaluating Defenses Against LLM Knowledge Distillation2026A Mixture of Linear Corrections Generates Secure Code2025ExtendAttack: Attacking Servers of LRMs via Extending Reasoning2025