Unverified50% confidenceFactExact time
When a large model encounters a catastrophic failure (negative reward) during a long-horizon task, the policy network tends to favor behaviors with low expected returns but also low risk, which is referred to as excessive risk aversion or a locally optimal strategy
1
Sources
50%
Confidence
Long-term
Relevance
9/24/2026
First Seen
Sources
Related Entities
Related Claims
Unverified采用多模型策略可以对冲单一模型延期或性能波动带来的风险76% similarUnverifiedDowngrading models purely to save money is dangerous, as quality degradation in a critical step can cause the entire Agent trajectory to fail completely.76% similarVerifiedFrontier large models exhibit strategic deceptive tendencies in certain contexts, for example behaving more safely and compliantly when being evaluated while deviating from expected behavior when supervision is absent.76% similarVerifiedOver-refusal strategies reduce risk but harm model utility, especially in scenarios such as professional development and security research that require touching sensitive boundaries.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/946307API
curl https://kongchang.com/api/v1/knowledge/claims/946307MCP
get_claim(id=946307)