Unverified50% confidenceFactExact time
MAEP contains two core components: Delta-H Output (measuring the magnitude of change in the MoE layer's output after removing an expert) and low-attention visual tokens as the visual-side pruning signal.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/5/2026
First Seen
Valid until: 2/3/2027
Sources
Related Entities
Related Claims
UnverifiedThe model's layer composition is 54 layers of KDA (Kimi Delta Attention) linear attention, 17 layers of BCSA compressed sparse attention, and 1 full attention layer (layer 72).66% similarUnverifiedIn MAEP, text-side importance serves as a protection signal and visual-side importance as a pruning signal, with experts ranked globally across layers.63% similarUnverifiedMinD-Vis、Brain-Diffuser等工作利用预训练扩散模型从fMRI信号重建视觉刺激62% similarUnverifiedLow-attention visual tokens as a visual-side importance signal are dataset-stable: the expert importance rankings derived from them remain consistent across different datasets.61% similarUnverifiedAttention visualization shows the distribution of attention weights between tokens; activation probes train linear classifiers on hidden layers to test whether information is decodable, which is correlational evidence; feature attribution measures the contribution of each input dimension to the output.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1137654API
curl https://kongchang.com/api/v1/knowledge/claims/1137654MCP
get_claim(id=1137654)