Unverified50% confidenceEventExact time
Within six weeks after tw-formosa-bench was released, frontier models broke through its calibrated difficulty window; on the public leaderboard, 22 of 108 models exceeded the benchmark's ceiling, with the best reaching 91.6%.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/15/2026
First Seen
Valid until: 10/29/2026
Sources
Related Entities
Related Claims
UnverifiedAIME has only 30 samples; claiming superiority by bolding a score of 80 against a baseline of 79 after running 4 seeds is statistically untenable.65% similarUnverifiedBenchmark overfitting is a core problem in current large model evaluation; some models achieve pass rates above 90% on HumanEval but perform far below expectations in real codebases.64% similarUnverified41个完整测试套件中有20个拿到满分100%，该基准已无法区分这些领先模型64% similarUnverifiedThe question set of tw-formosa-bench was filtered against a frozen pool of seven models so that their scores were kept at the passing margin (46.9%–74.6%, median about 54%).62% similarUnverifiedIn single-day changes (2026-10-12→2026-10-13), 2 models entered, 41 no longer appeared, and 201 had ranking changes; the real movement is at the edges, not the top.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1038585API
curl https://kongchang.com/api/v1/knowledge/claims/1038585MCP
get_claim(id=1038585)