paper-with-me

홈 › Papers

DeferMem: Query-Time Evidence Distillation via Reinforcement Learning for Long-Term Memory QA

2026-05-21 · Jianing Yin, Tan Tang arxiv

Large language model (LLM) agents still struggle with long-term memory question answering, where answer-supporting evidence is often scattered across long conversational histories and buried in substantial irrelevant content. Existing memory systems typically process memory before future queries are known, then retrieve the resulting units based on similarity rather than their utility for answering the query. This workflow leaves downstream answerers to denoise retrieved candidates and reconstruct query-specific evidence. We present DeferMem, a long-term memory framework that decouples this problem into high-recall candidate retrieval and query-conditioned evidence distillation. DeferMem uses a lightweight segment-link structure to organize raw history and retrieve broad candidates at query time. It then applies a memory distiller trained with DistillPO, our reinforcement learning algorithm for distilling the high-recall but highly noisy candidates into a set of faithful, self-contained, and query-conditioned evidence. DistillPO formulates post-retrieval evidence distillation as a structured action comprising message selection and evidence rewriting. It optimizes this action with a decomposed-and-gated reward pipeline and structure-aligned advantage assignment, gating reward components from validity to quality checks while exposing task-level correctness feedback early and assigning each reward to its responsible output span. On LoCoMo and LongMemEval-S, DeferMem surpasses strong baselines in QA accuracy and memory-system efficiency, achieving the highest QA accuracy with the fastest runtime and zero commercial-API token cost for memory operations.

📄 PDF Abstract BibTeX arXiv:2605.22411

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement LearningQuestion Answering

Similar Papers 제목 키워드 기반

DeepRetrieval: Hacking Real Search Engines and Retrievers with Large Language Models via Reinforcement Learning

2025-02-28 · Pengcheng Jiang, Jiacheng Lin, Lang Cao, Runchu Tian 외

Information retrieval systems are crucial for enabling effective access to large document collections. Recent approaches have leveraged Large Language Models (LLMs) to enhance retrieval performance through query augmenta…

Information Retrievalreinforcement-learningReinforcement LearningReinforcement Learning (RL)+1

DEEPRUBRIC: Evidence-Tree Rubric Supervision for Efficient Reinforcement Learning of Deep Research Agents

2026-06-15 · Minghang Zhu, Chuyang Wei, Junhao Xu, Yilin Cheng 외 arxiv

Deep research agents synthesize long-form reports by searching and reasoning over retrieved evidence. Reinforcement learning with rubric-based rewards improves these agents by optimizing them against checkable criteria t…

Reinforcement Learning

ViCuR: Visual Cues as Recoverable Privilege for Multimodal On-Policy Distillation

2026-06-04 · Kanghui Tian, Siyuan Liu, Ziang Yan, Sheng Xia 외 arxiv

On-policy distillation (OPD) improves reasoning by training a student on trajectories sampled from its own policy under supervision from a teacher. In multimodal reasoning, a common extension is to use a privileged teach…

Multimodal Reasoning

Context-Picker: Dynamic context selection using multi-stage reinforcement learning

2025-12-16 · Siyuan Zhu, Chengdong Xu, Kaiqiang Ke, Chao Yu arxiv

In long-context question answering, selecting the appropriate scope of context for a query remains a key and unresolved challenge. Insufficient context can lead to missing essential information, whereas excessive context…

Reinforcement LearningQuestion Answering

V-Zero: Answer-Label-Free On-Policy Distillation with Contrastive Evidence Gating for Fine-Grained Visual Reasoning

2026-06-24 · Haoxiang Sun, Zhihang Yi, Langxuan Deng, Yuhao Zhou 외 arxiv

Fine-grained visual reasoning requires multimodal large language models (MLLMs) to identify task-relevant visual evidence and ground their reasoning in local image regions. Existing agentic methods typically rely on rein…

Reinforcement LearningVisual Reasoning