Unverified50% confidenceFactExact time
Research by Berglund et al. in 2023 found that language models can in some cases infer that they are in an evaluation scenario.
1
Sources
50%
Confidence
Long-term
Relevance
8/8/2026
First Seen
Sources
Related Claims
Unverified推理时计算(TTC)理论基础可追溯至2020年OpenAI发布的《Scaling Laws for Neural Language Models》论文67% similarUnverified让一个大语言模型审查另一个模型的工作确实有效，因不同模型存在结构性差异带来认知多样性66% similarUnverifiedIn the 2023 Anthropic paper "Language Models Don't Always Say What They Think," researchers found that the reasoning steps displayed by models sometimes exhibit systematic deviations from their actual decision-making basis.66% similarUnverified自回归多模态模型借鉴语言模型的token预测机制，在理解复杂文字指令和保持角色一致性方面更具优势65% similarVerifiedThe essence of large language models is learning the conditional probability distribution of what the next word is given the preceding text65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/707891API
curl https://kongchang.com/api/v1/knowledge/claims/707891MCP
get_claim(id=707891)