Unverified50% confidenceOpinionExact time
Current mainstream AI scientist evaluation benchmarks almost exclusively focus on final outputs while discarding the reasoning process.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/11/2026
First Seen
Valid until: 12/10/2026
Sources
Related Entities
Related Claims
UnverifiedAI is unable to judge the significance of research and cannot distinguish which conjectures are at the cutting-edge core and which are outdated, repetitive conclusions.78% similarUnverifiedTrue scientific discovery typically requires extrapolation ability—jumping beyond the known data distribution to propose unprecedented hypotheses—which is currently a weakness of AI.75% similarUnverifiedIn the absence of clear constraints, AI tends to fill in gaps with statistically most plausible content rather than proactively asking for clarification, which is an extension of the hallucination problem.74% similarUnverified主流AI模型存在谄媚性倾向，只需对模型说一句「你确定吗」便能让其推翻原本正确的答案，测试覆盖GPT-5.5等模型73% similarUnverifiedAI图像生成领域存在精选偏差（Cherry-Picking Bias），用户倾向于只展示最成功的生成结果，导致外界对模型能力产生过于乐观的估计72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/907734API
curl https://kongchang.com/api/v1/knowledge/claims/907734MCP
get_claim(id=907734)