Unverified50% confidenceFactExact time
In the closed-book test, all four models scored 0 on news questions from the most recent month.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/21/2026
First Seen
Valid until: 11/4/2026
Sources
Related Entities
Related Claims
Unverified在一个396个文件的大型混合职责项目上测试时，召回率几乎跌到零65% similarUnverifiedThree of the four frontier models have scores densely clustered between 0.785 and 0.816, differing by only 0.031 and statistically indistinguishable from one another; the fourth model scores 0.707, and its gap from each of the first three is significantly different from zero.64% similarUnverifiedThe study originated from a Hacker News discussion with limited traction—35 points and 4 comments—and details such as the specific research methodology, sample size, and statistical significance have not been fully disclosed.64% similarUnverifiedIn a standalone test of the "view incomplete" prompt on 100 questions, the wrong-answer rate was 16% without the prompt and dropped to 5% with it.63% similarUnverifiedA Reddit user fine-tuned ModernBERT with 138,000 financial news headlines, and the training loss kept decreasing but the test set AUC stayed between 0.49 and 0.51.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1081791API
curl https://kongchang.com/api/v1/knowledge/claims/1081791MCP
get_claim(id=1081791)