Unverified50% confidenceBenchmarkExact time
When doing 50% pruning on Qwen3-4B, CACTUS raised the compressed-dense per-sample gradient cosine similarity on WikiText-2 from 0.670 to 0.755.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/14/2026
First Seen
Valid until: 1/12/2027
Sources
Related Entities
Related Claims
UnverifiedOn Qwen3-4B, after pruning MLP layers to 2:4 sparsity and adding rank-64 correction, CACTUS combined with LoGRA still maintains 0.711 cosine alignment with the C4 gradients of dense LoGRA after projection.80% similarUnverifiedWhen doing 50% pruning on Qwen3-4B, CACTUS reduced the gap between C4 perplexity and the dense model by 45%.73% similarUnverifiedUsing the Qwen3-27B model on the wikitext-2 dataset with fp32 reference logits as a baseline for KL divergence measurement, the median KLD dropped from 0.0023 to 0.000001, an approximately 2300-fold improvement in precision.63% similarUnverifiedQwen3覆盖从0.6B到235B的完整参数矩阵60% similarVerifiedThe reviewer believes that Qwen3.8 27B is stronger than the previous generation Qwen3.7 in almost all aspects of coding capability.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1022897API
curl https://kongchang.com/api/v1/knowledge/claims/1022897MCP
get_claim(id=1022897)