Unverified50% confidenceEventExact time
Meta研究团队在2024年展示了利用LLM生成Lean证明候选、再由Lean验证器确认正确性的流水线，成功证明了数学竞赛级别的问题
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/10/2026
First Seen
Valid until: 11/8/2026
Sources
Related Claims
UnverifiedMultiple studies in 2024 showed that LLMs exhibit a significant "self-consistency bias," tending to give overly favorable evaluations when asked to assess content they themselves generated.65% similarUnverifiedSnell等人在2024年发表研究《Scaling LLM Test-Time Compute Optimally》，量化了将算力从预训练转移到推理时搜索与验证的权衡关系63% similarVerifiedIn 2023, OpenAI published the paper "Let's Verify Step by Step," which systematically demonstrated the superiority of process reward models (PRM) in mathematical reasoning tasks.63% similarUnverified2021年的Chain-of-Thought Prompting论文证明，通过在提示词中加入推理步骤示例，模型的数学推理准确率可提升数十个百分点62% similarUnverified「LLM-as-Judge」范式最早在2023年被系统性提出，利用大语言模型本身的评估能力对生成内容进行质量打分61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/721298API
curl https://kongchang.com/api/v1/knowledge/claims/721298MCP
get_claim(id=721298)