Unverified50% confidenceFactExact time
When MAEM reaches a fixed capacity limit, it merges the least-used experts via the Git Re-Basin alignment technique, and distills the merged experts into a single unified fused expert using topology-aware pseudo-data synthesized from the MAE distributions.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/15/2026
First Seen
Valid until: 1/13/2027
Sources
Related Entities
Related Claims
Unverified选择性量化思路与LLM领域的GPTQ、AWQ等量化方法一脉相承67% similarUnverifiedThe idea of mixed-precision routing is consistent with the evolution of GPTQ and AWQ frameworks in the LLM quantization domain—namely, that understanding the activation distribution of each layer and dynamically allocating precision budgets is key to making low-bit quantization usable.66% similarUnverifiedMAEM is a replay-free, capacity-controllable dynamic mixture-of-experts framework for class-incremental graph learning.64% similarUnverifiedThe research team built three tools: AIPW-Fusion for estimating the ATE, and DR-Fusion and R-Fusion for estimating the CATE.64% similarUnverifiedMIT团队提出的AWQ通过分析激活值分布保护重要权重，在INT4量化下实现接近FP16的推理精度63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1033030API
curl https://kongchang.com/api/v1/knowledge/claims/1033030MCP
get_claim(id=1033030)