Unverified50% confidenceOpinionExact time
LLMs perform reliably in high-recall screening tasks such as title-abstract screening and full-text screening, but show a marked decline in evidence extraction tasks and still require human leadership.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedBecause LLM training data typically comes from large-scale web scraping and many benchmark questions are publicly available online, benchmark contamination is an almost unavoidable systemic problem65% similarUnverifiedLLM生成的虚假学术引用往往具有极高的可信外观，作者名、期刊名、年份、标题都遵循合理的分布，但组合结果可能从未真实存在65% similarUnverifiedDecomposed-subdomain walking is still a relatively frontier research direction, community discussion and independent verification are not yet sufficient, and specific implementation details and performance benchmark data await corroboration from the original paper or code.64% similarUnverifiedThe vast majority of LLM evaluation tools only score answers after they have already been presented to the user, and cannot intercept erroneous outputs in real time.64% similarUnverifiedLLM-based code auditing has difficulty performing precise state tracking across files and cross-function call chains, and cannot actually execute code to discover runtime vulnerabilities the way fuzz testing can.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/893428API
curl https://kongchang.com/api/v1/knowledge/claims/893428MCP
get_claim(id=893428)