Unverified50% confidenceBenchmarkExact time
The BF16 precision model's baseline accuracy was 74%, which improved to 84% after applying the penalty, with reasoning tokens reduced by 19.4%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/27/2026
First Seen
Valid until: 12/26/2026
Sources
Related Entities
Related Claims
UnverifiedIn the experiment, unit completion accuracy increased from 72.4% to 84.9%, a gain of 12.5 percentage points.76% similarUnverifiedWhen models were provided with complete company records, strict accuracy actually dropped from 41% to 18%, while recall rose and precision collapsed.72% similarUnverified4-bit quantization compared to FP16 typically results in a 0.5–2% drop in benchmark accuracy.72% similarVerifiedIf the accuracy per step is 95%, after 20 steps the overall accuracy will drop to approximately 36%.71% similarUnverified从BF16到FP8，内存占用减半，但数值表示精度大幅下降71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/954325API
curl https://kongchang.com/api/v1/knowledge/claims/954325MCP
get_claim(id=954325)