1
Robust Reasoning Benchmark
AIME 2024数学题经13种文本扰动，测试大模型推理鲁棒性，揭示依赖格式的短板
arXiv:2604.08571v2 Announce Type: replace-cross Abstract: While Large Language Models (LLMs) achieve high performance on standard mathematical benchma…
2
Open AI solves a 1946 Erdős problem
OpenAI用AI破解1946年数学难题，见证大模型推理能力新突破
Article URL: https://twitter.com/openai/status/2057176201782075690 Comments URL: https://news.ycombinator.com/item?id=48215185 Points: 3 # Comments: 0
3
Soohak: A Mathematician-Curated Benchmark for Evaluating Research-level Math Capabilities of LLMs
数学家精选的Soohak基准测试，专攻LLM科研级数学推理能力，挑战最高阶思维极限
arXiv:2605.09063v2 Announce Type: replace Abstract: Following the recent achievement of gold-medal performance on the IMO by frontier LLMs, the commun…
4
Implicit Hierarchical GRPO: Decoupling Tool Invocation from Execution for Tool-Integrated Mathematical Reasoning
将工具调用与执行解耦，提出隐式层次化GRPO框架，显著提升数学推理中的工具集成效率与泛化能力。
arXiv:2605.18500v1 Announce Type: new Abstract: Large language models (LLMs) have increasingly leveraged tool invocation to enhance their reasoning ca…
5
Lean Meets Theoretical Computer Science: Scalable Synthesis of Theorem Proving Challenges in Formal-Informal Pairs
结合Lean与理论计算机科学，可规模生成形式-非形式配对的定理证明挑战，助力AI数学推理研究。
arXiv:2508.15878v2 Announce Type: replace-cross Abstract: Formal theorem proving (FTP) has emerged as a critical foundation for evaluating the reasoni…
6
Beyond Benchmarks: MathArena as an Evaluation Platform for Mathematics with LLMs
针对静态基准测试的局限，推出MathArena平台，用动态、可扩展的评估助力LLM数学能力衡量
arXiv:2605.00674v2 Announce Type: replace Abstract: Large language models (LLMs) are becoming increasingly capable mathematical collaborators, but sta…