[KongchangAI]
Paper加权状态汤

Weighted State Souping for Efficient Long-Context Retrieval

An academic paper on long-context retrieval that proposes weighted state souping. By assigning weights to each independently encoded chunk state when combining context, it improves the retrieval performance of SSMs under crowded contexts.

Timeline (last 90 days)

Nov 11

The paper "Weighted State Souping for Efficient Long-Context Retrieval" was accepted to the PALM Workshop at NeurIPS 2026.

Unverified50%

All Facts (1)

Source Articles