Unverified50% confidenceFactExact time
The two hard zeros for graph retrieval in the benchmark were because the grader required exact wording while the model paraphrased; the README documents every point lost.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/18/2026
First Seen
Valid until: 11/1/2026
Sources
Related Entities
Related Claims
UnverifiedThe 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.76% similarUnverifiedThe benchmark produces two numbers for each model that are never merged: a task score and a false-confidence rate (the proportion of cases that should have been escalated but were answered anyway).74% similarUnverified困惑度是测试集上每个token平均负对数似然的指数，值越低代表模型对语料的预测越准确74% similarUnverifiedIn differential testing, a non-zero curl exit should fail the case outright, because two identical errors look like a match to the diff.73% similarUnverifiedThe core of evaluation is scoring, typically expressing the quality of results using a continuous value from 0 to 1.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1062578API
curl https://kongchang.com/api/v1/knowledge/claims/1062578MCP
get_claim(id=1062578)