Unverified50% confidenceEventExact time
A cropping change that produced no change at all on the tuning package caused that page's accuracy to plummet from 94% to 86% on two check packages, and the author rolled back the change.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/18/2026
First Seen
Valid until: 1/16/2027
Sources
Related Entities
Related Claims
UnverifiedAfter removing redundant content such as repeated headers and footers, the token count drops by approximately 40% to 65%.73% similarUnverifiedA gating accuracy of 58% reported in the paper was mistakenly treated as end-to-end quality, whereas the true end-to-end quality is 98%.71% similarUnverifiedIn the control group where the answer key was removed from the prompt, the match rate plummeted from 63% to 1% (1/75).70% similarUnverifiedLooking only at the two check packs, the fully accurate rate rose from 94.4% to 96.0%.70% similarUnverifiedAfter removing the candidate filtering component, FEVER accuracy dropped to 76.24%.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1063717API
curl https://kongchang.com/api/v1/knowledge/claims/1063717MCP
get_claim(id=1063717)