Unverified50% confidenceFactExact time
In off-policy settings, importance sampling weights must be applied on top, where these weights are the ratio of the current policy's probability to the behavior policy's probability, and extreme values can greatly amplify variance.
1
Sources
50%
Confidence
Long-term
Relevance
9/20/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOptimizing for benchmarks carries risks, as high scores may only reflect a model's adaptation to specific question types rather than a genuine improvement in general capabilities.78% similarUnverified在大多数复杂优化问题中，贪心策略会陷入局部极值而非全局极值74% similarUnverifiedA model's 'raw capability level' and its 'actual deployment risk' are two concepts that need to be considered separately.73% similarUnverifiedModel selection is essentially a multi-objective optimization problem: sometimes the lowest price is needed, sometimes the fastest response, and sometimes specific capabilities.73% similarUnverifiedTraditional risk management theory holds that increasing the number of independent decision-making entities can diversify risk, but this mechanism fails when decision-making entities are highly correlated.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/937385API
curl https://kongchang.com/api/v1/knowledge/claims/937385MCP
get_claim(id=937385)