Unverified50% confidenceFactExact time
LTT does not rely on the premise that the probabilistic scores are exactly accurate, but instead provides a finite-sample risk guarantee: as long as the set of candidate thresholds is fixed in advance, the probability that the threshold selected by LTT exceeds the target risk on the true distribution is strictly controlled.
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
Unverified安全对齐本质是在模型概率分布上施加偏置，而非从根本上消除有害能力74% similarUnverifiedThe coverage guarantee of conformal prediction is a frequentist marginal guarantee that holds on the overall test set but does not guarantee coverage on any arbitrary subset.71% similarUnverifiedBefore deploying LLMs in high-stakes professional settings, the underlying model reasoning may be unreliable beneath a surface of high accuracy, making rigorous testing of contextual adaptability indispensable.71% similarUnverifiedA high-probability upper bound means that with probability at least 1−δ the regret does not exceed some expression in terms of the number of rounds T, which is stronger than a bound in the expectation sense.69% similarUnverifiedRaising a model's lower bound (reducing the frequency of stupid mistakes) does not increase potential risk, whereas pushing up the upper bound (improving peak capability) necessarily comes with a rise in potential danger.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1037405API
curl https://kongchang.com/api/v1/knowledge/claims/1037405MCP
get_claim(id=1037405)