Unverified50% confidenceOpinionExact time
The unique value of DEL data lies in the fact that negative samples come from real experimental non-binding results and cover a broad chemical space; each round of selection amounts to an empirical test of the model's hypotheses, offering a test of true generalization closer to reality than synthetic benchmarks.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
Unverified对比学习中的推远操作只要求负样本相似度低于正样本，并不要求达到-1的极端值68% similarUnverifiedThe comparative experiment did not disclose specific dataset sizes, evaluation metrics, or details of comparison fairness, all of which affect the generalizability of its conclusions.67% 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67% similarUnverifiedOn the strongest models, the differences between experimental groups fall within confidence intervals, and researchers characterized the results as directional patterns and soundness properties rather than proven performance lower bounds66% similarUnverifiedRecording negative samples where 'a recommended solution did not apply' is just as important as recording successes; otherwise the model will fall into statistical bias, mistaking high-frequency solutions for high-accuracy ones66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1031868API
curl https://kongchang.com/api/v1/knowledge/claims/1031868MCP
get_claim(id=1031868)