Unverified50% confidenceFactExact time
The formula for relative WER improvement is: (WER_SFT - WER_GRPO) / WER_SFT × 100%.
1
Sources
50%
Confidence
Long-term
Relevance
10/26/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe match score calculation formula is: score = (strong*1.0 + partial*0.6 + transferable*0.4) / total60% similarUnverifiedOn the MATH mathematical reasoning benchmark, JOLT achieved the same accuracy as GRPO, but processed only about one-sixth as many training tokens as GRPO.59% similarVerifiedThe core formula of RRF is RRF_score(d) = Σ 1 / (k + rank_i(d)), where k is a smoothing constant commonly set to 60.59% similarUnverifiedIn the study, the Full graph method achieved an SHD of 135 and an F1-score of 0.73, performing better in terms of structural distance.59% similarUnverifiedCFG的数学公式为ε_guided = ε_uncond + w × (ε_cond - ε_uncond)，典型CFG值范围是5-1559% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1098815API
curl https://kongchang.com/api/v1/knowledge/claims/1098815MCP
get_claim(id=1098815)