Unverified50% confidenceFactExact time
In stochastic sampling decoding, a high-confidence predictive distribution can still sample low-probability tokens, so testing the distribution alone cannot guarantee the reliability of the tokens actually written.
1
Sources
50%
Confidence
Long-term
Relevance
11/4/2026
First Seen
Sources
Related Entities
Related Claims
Unverified在机器学习多种子实验中重复次数通常只有5到10次，样本量极小，正态性假设难以验证，因此 Wilcoxon 检验往往是更保守稳健的选择75% similarUnverifiedIf traditional jailbreak evaluation uses only greedy decoding, it yields a deterministic single output that cannot reflect the model's true behavior distribution under stochastic sampling.74% similarUnverifiedThe verification process in traditional speculative decoding accepts draft-proposed tokens with probability min(1, p(x)/q(x)), and upon rejection resamples from the corrected distribution norm(max(0, p(x)-q(x))), which can be proven to produce a token distribution identical to directly sampling from the target model (lossless).74% similarUnverifiedConfidence values typically originate from the model's softmax probability distribution over each option — a large probability gap yields a high confidence value, while a uniform distribution yields a low one.74% similarUnverifiedDue to sampling variance, each test message should be run twice; a rule that holds with only 50% probability is effectively no rule at all.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1134050API
curl https://kongchang.com/api/v1/knowledge/claims/1134050MCP
get_claim(id=1134050)