Unverified50% confidenceBenchmarkExact time
When using the multi-agent verification process, MARCH's accuracy was suppressed to 4.4%, whereas when the same model was directly asked which solution was better, accuracy reached 43.7%.
1
Sources
50%
Confidence
Long-term
Relevance
9/28/2026
First Seen
Sources
Related Entities
Related Claims
Unverified4-bit quantization compared to FP16 typically results in a 0.5–2% drop in benchmark accuracy.77% similarUnverifiedWhen models were provided with complete company records, strict accuracy actually dropped from 41% to 18%, while recall rose and precision collapsed.76% similarVerifiedIf the accuracy per step is 95%, after 20 steps the overall accuracy will drop to approximately 36%.75% similarVerifiedResearch indicates that at Q4 precision and above, performance loss is within 1–3% for most tasks.75% similarUnverified在标准行人检测数据集（如MOTChallenge）上训练的模型，在俯视视角下的检测精度（mAP）通常下降40%-60%75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/955489API
curl https://kongchang.com/api/v1/knowledge/claims/955489MCP
get_claim(id=955489)