Unverified50% confidenceOpinionExact time
Having another large model read the paper cannot substitute for genuine verification, as mutual evaluation between models is prone to failing on details.
1
Sources
50%
Confidence
Long-term
Relevance
9/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedBecause large models themselves cannot strictly distinguish between instructions and data, prompt injection is fundamentally very difficult to completely eliminate.74% similarVerifiedLarge language models tend to solve problems using general reasoning while lacking strict adherence to domain-specific rules and self-verification capabilities.73% similarUnverifiedThe researchers interpreted the gap between selection and generation as an 'evaluation format gap,' rather than direct evidence that the model's lexical knowledge is intact.72% similarUnverifiedPure vector retrieval can only measure semantic proximity and cannot establish explicit logical associations between entities, making it difficult to handle cross-document comprehensive reasoning problems.72% similarUnverifiedLonger prompts are not necessarily better; irrelevant filler should be removed, retaining only the information the model needs to be told.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/946343API
curl https://kongchang.com/api/v1/knowledge/claims/946343MCP
get_claim(id=946343)