Unverified50% confidenceBenchmarkExact time
On 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.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/14/2026
First Seen
Valid until: 1/12/2027
Sources
Related Entities
Related Claims
UnverifiedWhen doing 50% pruning on Qwen3-4B, CACTUS raised the compressed-dense per-sample gradient cosine similarity on WikiText-2 from 0.670 to 0.755.80% similarUnverifiedWhen doing 50% pruning on Qwen3-4B, CACTUS reduced the gap between C4 perplexity and the dense model by 45%.79% similarUnverifiedQwen3覆盖从0.6B到235B的完整参数矩阵68% similarUnverifiedQwen3 is available on Ollama in multiple sizes including 0.6B, 1.7B, 4B, and 8B65% similarUnverifiedQwen3-1.7B's faithfulness is 68% on 2-hop tasks but plummets to 30% on 6-hop tasks, whereas Qwen3-4B remains relatively stable across different hop counts.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1022900API
curl https://kongchang.com/api/v1/knowledge/claims/1022900MCP
get_claim(id=1022900)