Unverified50% confidenceBenchmarkExact time
When looking only at logprob, 17 cases (47%) appeared to successfully suppress PII; but with a multi-metric framework, only 1 case (2.8%) passed.
1
Sources
50%
Confidence
Long-term
Relevance
11/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe best-performing largest model can solve 79.6% of held-out test set problems, but only 55.3% on TranSGrid, dropping sharply to 15.8% on the hardest subset.65% similarUnverifiedIn 13% to 38% of correct answers, the model's reasons for eliminating distractors were clinically untenable, and confidence did not reflect this at all.64% similarUnverifiedAt a significance level of ε=0.09, 63% of nodules received a definitive single-label prediction, while the remaining 37% of cases were referred to clinicians for judgment.63% similarUnverifiedAggregate metrics (such as average scores) mask failures in long-tail scenarios; when 95% of routine cases perform excellently, the collapse of the remaining 5% of edge cases leaves no obvious mark on the overall score.63% similarUnverifiedAfter applying logit penalties, the number of reasoning tokens decreased by 11% to 19% across all precision formats.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1133781API
curl https://kongchang.com/api/v1/knowledge/claims/1133781MCP
get_claim(id=1133781)