Unverified50% confidenceBenchmarkExact time
MATH包含12,500道竞赛数学题按难度分五级，GPQA收录研究生级别谷歌难以直接找到答案的科学问题
1
Sources
50%
Confidence
Long-term
Relevance
7/10/2026
First Seen
Sources
「GPT-5.6」真相揭秘：虚假传闻与第三方充值陷阱辨析
bilibili小小阿白6667/9/2026
Related Claims
UnverifiedGPT-5 Codex的网络搜索功能生成的搜索查询质量极差，模型不擅长构造有效的搜索查询57% similarUnverifiedGoogle Deep Research在测试中直接遗漏了解说稿的输出，未完成全部任务56% similarUnverifiedPrisma生成的查询在某些复杂场景下可能不如手写SQL灵活，抽象层会带来学习成本和运行开销55% similarUnverified斯坦福大学研究者测试多个主流模型发现，措辞中性的医疗、法律或安全研究类查询也会因触发关键词匹配而被错误拒绝54% similarUnverifiedBecause LLM training data typically comes from large-scale web scraping and many benchmark questions are publicly available online, benchmark contamination is an almost unavoidable systemic problem54% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/463387API
curl https://kongchang.com/api/v1/knowledge/claims/463387MCP
get_claim(id=463387)