Unverified50% confidenceSolutionExact time
When fine-tuning a cross-encoder for classification, each sample must be constructed as a (text, label description) pair, with the label being a natural language description rather than an integer ID.
1
Sources
50%
Confidence
Long-term
Relevance
10/21/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe mention counts are computed by code based on keywords rather than asserted by a model, so the counting methodology can be verified.69% similarUnverified作者选择自定义标签系统而非CLIP等预训练文本编码器69% similarUnverifiedA controlled vocabulary is a predefined, limited and semantically clear set of exclusion reason labels rather than free text; its benefits are that exclusion records can be aggregated for analysis and compared across cases, and vague wording is avoided.68% similarUnverifiedGLiNER将实体识别转化为候选标签与文本片段之间的语义相似度匹配问题，可在推理时接受任意新标签而无需重新训练67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1083225API
curl https://kongchang.com/api/v1/knowledge/claims/1083225MCP
get_claim(id=1083225)