analiza aiqbench
SaliMory addresses the critical challenge of persistent memory in conversational agents by replacing raw context expansion with a cognitively structured framework that isolates supervision for distinct memory operations. This approach resolves credit assignment bottlenecks in multi-stage pipelines, resulting in a one-third reduction in memory-attributed failures and a doubling of the Good Personalization rate compared to state-of-the-art methods. The technique represents a significant methodological advance for building reliable, lifelong companion agents without degrading reasoning quality.
SaliMory: Orchestrating Cognitive Memory for Conversational Agents
arXiv:2606.04120v1 Announce Type: new Abstract: Conversational agents that serve as lifelong companions must maintain persistent memory across all interactions. However, simply expanding context windows with raw retrieval degrades reasoning quality, while training memory agents via standard reinforcement learning creates a severe credit assignment bottleneck in a multi-stage pipeline. To solve this, we introduce SALIMORY, a framework that trains a single language model to manage a cognitively-structured memory-spanning user facts, preferences, and working memory. By introducing a hierarchical stage-wise process reward and reward-decomposed contrastive refinement, SALIMORY provides isolated supervision for distinct memory operations (selective filtering, consolidation, and cue-driven recall) end-to-end. SALIMORY cuts memory-attributed failures by one-third, outperforms the state-of-the-art by over 10% in end-to-end accuracy, and more than doubles the Good Personalization rate.
Abstract:Conversational agents that serve as lifelong companions must maintain persistent memory across all interactions. However, simply expanding context windows with raw retrieval degrades reasoning quality, while training memory agents via standard reinforcement learning creates a severe credit assignment bottleneck in a multi-stage pipeline. To solve this, we introduce SALIMORY, a framework that trains a single language model to manage a cognitively-structured memory-spanning user facts, preferences, and working memory. By introducing a hierarchical stage-wise process reward and reward-decomposed contrastive refinement, SALIMORY provides isolated supervision for distinct memory operations (selective filtering, consolidation, and cue-driven recall) end-to-end. SALIMORY cuts memory-attributed failures by one-third, outperforms the state-of-the-art by over 10% in end-to-end accuracy, and more than doubles the Good Personalization rate.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2606.04120 [cs.CL]
(or arXiv:2606.04120v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2606.04120 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 139 cuvinte
Subiecte conexe
SaliMory: Orchestrating Cognitive Memory for Conversational Agents
arXiv · 27 aug. 2026
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · 17 sept. 2026
[AINews] OpenAI reports Navier-Stokes singularity find in 88 hours using Astra-next, roughly 10,000 agents and 130B tokens (>$40M), a contender for second ever Millennium Prize awarded
Latent Space · 9 sept. 2026
Biological Superintelligence
LessWrong · 30 iul. 2026