Unverified50% confidenceFactExact time
The true failure mode in verbatim-quotation tasks is disagreement over sentence-boundary judgment, not typo correction.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWord count errors are difficult to eliminate entirely through prompt improvements, because the problem lies in a fundamental mismatch between the nature of the task and the model's working mechanism, not in whether the wording is clear76% similarUnverifiedThe primary failure mode of early language models on intent understanding tasks was 'literal execution' rather than 'intent execution,' meaning they strictly followed literal instructions while ignoring the business outcomes the user actually expected.75% similarUnverified纯写死规则的意图识别方案泛化能力差，用户换一种说法就识别失败73% similarUnverifiedThe persuasiveness trap of large language models is essentially a calibration failure, as there is no positive correlation between output fluency and factual accuracy.71% similarUnverifiedUsing incorrect default capability advertisement values masks the true situation and leads to difficult-to-reproduce errors, whereas unknown is an explicit semantic state that callers can use for graceful degradation71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1000252API
curl https://kongchang.com/api/v1/knowledge/claims/1000252MCP
get_claim(id=1000252)