Unverified50% confidenceFactExact time
In WISD, the average improvement from pruning is higher than that from quantization, and is especially pronounced on small models.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Related Entities
Related Claims
UnverifiedOn math and code reasoning benchmarks, WISD improves average accuracy by 4.0 percentage points with quantization-based weakening and by 5.5 percentage points with pruning-based weakening.73% similarUnverifiedA larger base model, even after aggressive quantization, still outperforms a smaller-parameter model with light quantization in overall capability.63% similarUnverifiedAccording to the study's description, this processing noticeably improves the model's accuracy and consistency of judgment on samples involving binary payloads—the blind spot is partially filled.61% similarUnverifiedIn the extended trials on 26 models, numerical computation performance rises together with non-numerical performance.61% similarUnverifiedInjecting concept representation into the draft model of DFlash2 can improve the mean acceptance length by 4.17%.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114235API
curl https://kongchang.com/api/v1/knowledge/claims/1114235MCP
get_claim(id=1114235)