Unverified50% confidenceFactExact time
In the task of predicting final exam scores from midterm scores, a basic neural network using only students' own scores has high error.
1
Sources
50%
Confidence
Long-term
Relevance
10/26/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMany neural networks exhibit an overconfidence calibration bias, where output confidence scores remain close to 1 even when the error rate is high.71% similarUnverifiedTraditional benchmarks have become ineffective for evaluating model reasoning ability because almost every model may have seen the standard answers in its training data.66% similarUnverifiedThe SGUID study points out that in on-policy distillation, fewer than 25% of the retrieved skills can provide effective learning signals.65% similarUnverifiedMultiple independent studies in 2023 showed that some top models scored extremely high on standard mathematical reasoning tests but performed significantly worse on equivalent rephrased versions, indicating that the models learned test patterns rather than genuine reasoning ability.65% 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65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1101055API
curl https://kongchang.com/api/v1/knowledge/claims/1101055MCP
get_claim(id=1101055)