Memory Retrieval for Changing Preferences
Quick Take
This paper introduces a unified framework for memory retrieval in dialogue systems that adapts to changing user preferences by using Bayes factors to quantify the utility of historical interactions. Experiments demonstrate that this approach outperforms existing methods on long-context tasks, enhancing performance in scenarios where user preferences evolve.
Key Points
- Proposes a memory retrieval framework based on changing user preferences.
- Utilizes Bayes factors to assess the utility of historical dialogue turns.
- Outperforms embedding-based retrieval in long-context, preference-intensive tasks.
- Maintains competitive performance in low-density scenarios with semantic similarity.
- Addresses the limitations of heuristic retrieval signals in existing systems.
Article Content
From source RSS / original summaryarXiv:2606. 02976v1 Announce Type: new Abstract: Long-context dialogue systems must decide both when to access memory and which parts of the interaction history are relevant. Existing approaches typically rely on heuristic retrieval signals or always-on memory usage, failing to account for the changing and potentially inconsistent nature of user preferences. In this work, we propose a unified framework for memory access and selection based on changing preferences.
We formulate personalized memory retrieval as identifying which historical turns provide evidence about a user's latent preference state, rather than relying on surface-level semantic similarity. To this end, we quantify the utility of each memory turn using a Bayes factor, defined as the improvement in the model's likelihood of the reference response when the turn is included in context. This provides a principled measure of evidence strength and a unified signal for both memory access and selection.
By framing memory retrieval as utility estimation, the model learns to identify salient turns and regulate memory usage based on expected utility. Experiments on four heterogeneous memory benchmarks show that our approach outperforms existing embedding-based retrieval on long-context, preference-intensive tasks where modeling changing preferences is essential, while remaining competitive in low-density regimes where semantic similarity suffices.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?
The REFLECT benchmark reveals that current LLM judges are unreliable, achieving below 55% accuracy in evaluating reasoning and evidence use, highlighting the need for improved evaluation methods for deep research agents.