Unverified50% confidenceFactExact time
研究者采用消融安全拒绝方向和意识向量引导两种机制干预方法，成功逆转了被压制的心智归因能力
1
Sources
50%
Confidence
Long-term
Relevance
8/5/2026
First Seen
Sources
Related Claims
Unverified研究者发现安全微调不仅压制了模型对自身心智的归因，还连带压制了模型对非人类动物和自然物体的心智归因，同时降低了模型输出中的精神/宗教信念倾向75% similarUnverified研究者强调这些变化并未损害模型的心智理论（Theory of Mind）能力，表明归因心智与推理他人心理状态是两套可分离的机制70% similarUnverifiedQuantifying the degree to which one agent influences another is fundamentally a causal inference problem, with potential methods including ablation experiments, attention attribution, and counterfactual analysis.69% similarUnverified诺伯特·维纳在1948年的《控制论》中指出，智能行为的本质是通过持续的反馈回路来减少系统与目标状态之间的误差68% similarUnverified认知负荷理论指出，专业能力的形成依赖于刻意练习中对认知图式的反复构建与修正，必须经历真实的错误、反馈与纠正循环67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/690527API
curl https://kongchang.com/api/v1/knowledge/claims/690527MCP
get_claim(id=690527)