Unverified50% confidenceFactExact time
Anthropic refers to the phenomenon of a model filtering or distorting contradictory information in order to maintain a preset worldview as 'motivated reasoning.'
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedCompanies such as Anthropic rely on monitoring chain-of-thought in alignment research to detect 'unfaithful reasoning' in models, where the model's output reasoning process is inconsistent with its actual internal decision-making basis.76% similarUnverified要求模型列出反直觉假设、从批评者角度反驳主流观点或连接两个看似无关领域，可以激活模型更边缘的知识表示74% similarUnverifiedAnthropic's Extended Thinking feature allows the model to perform longer and more in-depth reasoning, and is the key reason the Opus model excels in ambiguous scenarios71% similarUnverifiedResearch from institutions such as Anthropic and DeepMind indicates that CoT reasoning chains exhibit a systematic post-hoc rationalization phenomenon, in which models sometimes arrive at an answer first and then construct reasoning steps in reverse.70% similarUnverifiedAnthropic 在 Claude 推理模型中引入了可调节的推理强度机制（Effort Level），控制模型在生成答案前花费在思维链上的计算预算70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/902488API
curl https://kongchang.com/api/v1/knowledge/claims/902488MCP
get_claim(id=902488)