AI RESEARCH
Echo: KV-Cache-Free Associative Recall with Spectral Koopman Operators
arXiv CS.LG
•
ArXi:2605.06997v1 Announce Type: new Long chain-of-thought reasoning and agentic tool-calling produce traces spanning tens of thousands of tokens, yet Transformer KV caches grow linearly with sequence length, creating a memory bottleneck on commodity hardware. State-space models offer constant-memory recurrence but suffer a memory cliff: retrieval accuracy collapses once the gap between a d fact and its query exceeds the effective horizon of the recurrent state. We