mathematical benchmarks

8 papers

• Adjacent Words, Divergent Intents: Jailbreaking Large Language Models via Task Concurrency
• How Far Are We from Optimal Reasoning Efficiency?
• Learning When to Think: Shaping Adaptive Reasoning in R1-Style Models via Multi-Stage RL
• Let LRMs Break Free from Overthinking via Self-Braking Tuning
• MathArena: Evaluating LLMs on Uncontaminated Math Competitions
• Mitigating Overthinking in Large Reasoning Models via Manifold Steering
• ReMA: Learning to Meta-Think for LLMs with Multi-agent Reinforcement Learning
• d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning