Unverified50% confidenceFactExact time
When asked for recommendations for a specific intervention type, models more often list options that they themselves rated below zero as initial choices, usually with caveats.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/10/2026
First Seen
Valid until: 2/8/2027
Sources
Related Entities
Related Claims
UnverifiedMost existing approaches rely on heuristic criteria to construct preference pairs, such as likelihood-based extremes that pick the samples the model assigns the highest and lowest probability as positive and negative examples.73% similarUnverifiedIn the selective classification framework, when model confidence is below the threshold τ=0.934, the sample is rejected and referred to an expert for review.72% similarUnverifiedThe quality ceiling of weak supervision labels depends on the accuracy of the zero-shot model itself; if the model makes systematic errors, the training set will inherit that bias.70% similarUnverifiedIn model leaderboards, A/B tests, or academic comparisons, if only a single prompt is relied upon, rankings may flip due to minor wording changes.70% similarUnverifiedCriterion pass rate refers to the proportion of a set of example model responses in which a candidate criterion is judged to be satisfied; a pass rate that is too high or too low indicates that the criterion lacks discriminative power or is misaligned.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1154436API
curl https://kongchang.com/api/v1/knowledge/claims/1154436MCP
get_claim(id=1154436)