Unverified50% confidenceFactExact time
After introducing retrieval, 20 of the 26 models improved, while 6 declined.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/3/2026
First Seen
Valid until: 2/1/2027
Sources
Related Entities
Related Claims
UnverifiedAmong 27 evaluation settings, distillation narrowed the group gap in 21 of them.72% similarUnverifiedAfter introducing the dispute mechanism, the same model's success rate became 20/20; the control scenario (an app in the foreground but no one touching it) still maintained 20/20 with zero model calls.71% similarUnverifiedWhen the 35B model switched to a wider context, its pass rate plummeted to 46/90, with 43 responses truncated due to a fixed output budget.69% similarUnverifiedThe developer cautions that the sample size of 18 matchups is insufficient to rank the models reliably.68% similarUnverified2024年多项研究发现，部分模型在去污染版本的评测集上得分下降幅度可达10-20个百分点68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1130213API
curl https://kongchang.com/api/v1/knowledge/claims/1130213MCP
get_claim(id=1130213)