Unverified50% confidenceFactExact time
The authors' reproduced scores for multilingual-e5-large-instruct and BGE-M3 deviate from published MTEB results by no more than 2.4 percentage points.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/10/2026
First Seen
Valid until: 2/8/2027
Sources
Related Entities
Related Claims
UnverifiedThe author uses a weighted scoring scheme: in the outcome dimension, Pass scores 1.0, Partial scores 0.5, and Fail scores 0; severity weights are 4 for Critical, 3 for High, 2 for Medium, and 1 for Low.65% similarUnverifiedAfter re-evaluating with non-overlapping anchors, the net advantage instead grew from +4.1pp to +5.5pp, but significance dropped from 2.4 standard errors to 1.4 standard errors; the inflation factor is about 2.4x, exactly equal to sqrt(912/163).64% similarUnverifiedo4-mini and gpt-5-thinking-mini have almost the same accuracy on SimpleQA, but the latter makes roughly three times fewer errors than the former.63% similarUnverifiedOn the Joyo-Kanji-Yomi benchmark, Ruby-OPD achieves an accuracy of 97.13%, a Target Phoneme Error Rate (Target PER) of 2.02%, a Sentence-level Phoneme Error Rate (Sentence PER) of 0.70%, and a Sentence-level Character Error Rate (Sentence CER) of 4.21% under raw text input.62% similarUnverifiedThe int4 build from Google's QAT checkpoint on Gemma 4 E2B is less than 0.5 points lower in accuracy than bf16 (0.678 vs 0.683), while the int8 build rounded directly from bf16 dropped 1.3 points.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1153283API
curl https://kongchang.com/api/v1/knowledge/claims/1153283MCP
get_claim(id=1153283)