Unverified50% confidenceBenchmarkExact time
When small models are made to do verbatim quotations, only 24% to 35% of the quotations pass verification against the source text.
1
Sources
50%
Confidence
Long-term
Relevance
10/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAfter implementing a process of downloading sources first and then letting the model write, 36 of 39 research citations were verbatim correct, and 95% of quotations in thinker biographies passed source verification.72% similarUnverifiedEven the best-performing models can only recover 30% of the evaluated evidence and 25% of the limitations descriptions.70% similarUnverifiedIn manually proofread samples, the character error rate was only 0.68% when the text was clearly legible, but approximately 45% of sampled passages were completely uncorrectable due to source quality issues.68% similarUnverifiedA developer reviewed 27 prompts accumulated over one year, of which 22 failed the reproducibility test (i.e., others using the prompts could not obtain the same output as the author).67% similarUnverifiedThe study compared on four models, two in-domain corpora, and three held-out corpora, and the results show that only ASC simultaneously reduces the 'authored sentence' error rate and increases supported sentences.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1049790API
curl https://kongchang.com/api/v1/knowledge/claims/1049790MCP
get_claim(id=1049790)