Unverified50% confidenceCautionExact time
The handling of negation by embedding models is a systematic flaw, because mainstream training objectives (masked language modeling, contrastive learning) do not explicitly require the model to distinguish between positive and negative semantics, causing 'like' and 'dislike' to be encoded at nearby positions.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe pretraining corpora of the vast majority of models are closed-source, forcing membership inference research to rely on heuristic methods to construct members and non-members, which may introduce systematic bias.78% similarUnverifiedInner misalignment refers to the phenomenon where the goals a model learns during training are not fully identical to the training objectives.77% similarUnverifiedCurrent mainstream large language models exhibit significant deficiencies in both confidence calibration and selective abstention, tend to be overconfident, and when faced with questions outside the training distribution, tend to produce fluent but incorrect answers.76% similarUnverifiedNative large models only possess knowledge from their training period and often produce incorrect answers about events that occurred after training.75% similarUnverifiedModel routing is essentially choosing among capabilities trained by others and cannot change the judgment or knowledge boundaries of each model itself, so it will still hit a ceiling.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/995602API
curl https://kongchang.com/api/v1/knowledge/claims/995602MCP
get_claim(id=995602)