arXiv:2601.07994v1 Announce Type: new 
Abstract: Large Language Models (LLMs) often exhibit increased response latency and degraded answer quality as dialogue length grows, making effective context management essential. However, existing methods rely on extra LLM calls to build memory or perform offline memory construction without considering the current user utterance, which can introduce inefficiencies or disrupt conversational continuity. We introduce DyCP, a lightweight context management method that dynamically segment and retrieve relevant memory at query time. It preserves the sequential structure of dialogue without predefined topic boundaries and supports efficient, adaptive context retrieval. Across three long-form dialogue benchmarks, LoCoMo, MT-Bench+, and SCM4LLMs, and multiple LLMs, DyCP consistently improves answer quality while reducing response latency. We also examine the gap between modern LLMs' expanded context windows and their actual long-context processing capacity, highlighting the continued importance of effective context management.

تم تقديم طريقة جديدة تُدعى DyCP (Dynamic Context Pruning) لتحسين أداء نماذج اللغة الكبيرة (LLMs) في الحوارات الطويلة من خلال تقسيم واسترجاع الذاكرة ذات الصلة ديناميكيًا في وقت الاستعلام، مما يحسن جودة الإجابات ويقلل من زمن الاستجابة.

Se ha introducido un nuevo método llamado DyCP (Dynamic Context Pruning) para mejorar el rendimiento de los Modelos de Lenguaje Grande (LLMs) en diálogos de larga duración, segmentando y recuperando dinámicamente la memoria relevante en el momento de la consulta, mejorando así la calidad de las respuestas y reduciendo la latencia.

Une nouvelle méthode appelée DyCP (Dynamic Context Pruning) a été introduite pour améliorer la performance des modèles de langage de grande taille (LLMs) dans les dialogues longs en segmentant et en récupérant dynamiquement la mémoire pertinente au moment de la requête, améliorant ainsi la qualité des réponses tout en réduisant la latence.

A new method called DyCP (Dynamic Context Pruning) has been introduced to enhance the performance of Large Language Models (LLMs) in long-form dialogues by dynamically segmenting and retrieving relevant memory at query time, improving answer quality while reducing response latency.

DYCP: Dynamic Context Pruning for Long-Form Dialogue with LLMs

arXiv:2601.08160v1 Announce Type: new 
Abstract: Agentic memory systems have become critical for enabling LLM agents to maintain long-term context and retrieve relevant information efficiently. However, existing memory frameworks suffer from a fundamental limitation: they perform exhaustive retrieval across the entire storage layer regardless of query characteristics. This brute-force approach creates severe latency bottlenecks as memory grows, hindering real-time agent interactions. We propose SwiftMem, a query-aware agentic memory system that achieves sub-linear retrieval through specialized indexing over temporal and semantic dimensions. Our temporal index enables logarithmic-time range queries for time-sensitive retrieval, while the semantic DAG-Tag index maps queries to relevant topics through hierarchical tag structures. To address memory fragmentation during growth, we introduce an embedding-tag co-consolidation mechanism that reorganizes storage based on semantic clusters to improve cache locality. Experiments on LoCoMo and LongMemEval benchmarks demonstrate that SwiftMem achieves 47$\times$ faster search compared to state-of-the-art baselines while maintaining competitive accuracy, enabling practical deployment of memory-augmented LLM agents.

تم تقديم SwiftMem كنظام ذاكرة وكيل واعٍ للاستعلامات، مصمم لتحسين كفاءة وكلاء نماذج اللغة الكبيرة (LLM) من خلال تمكين الاسترجاع دون خطي عبر تقنيات الفهرسة المتخصصة. يعالج هذا النظام قيود الأطر الحالية للذاكرة التي تعتمد على طرق الاسترجاع الشاملة، والتي يمكن أن تؤدي إلى مشاكل كبيرة في الكمون مع توسع تخزين الذاكرة.

SwiftMem se ha presentado como un sistema de memoria agentica consciente de las consultas, diseñado para mejorar la eficiencia de los agentes de modelos de lenguaje de gran tamaño (LLM) al permitir una recuperación sub-lineal a través de técnicas de indexación especializadas. Este sistema aborda las limitaciones de los marcos de memoria existentes que dependen de métodos de recuperación exhaustiva, que pueden llevar a problemas significativos de latencia a medida que se expande el almacenamiento de memoria.

SwiftMem a été présenté comme un système de mémoire agentique conscient des requêtes, conçu pour améliorer l'efficacité des agents de modèles de langage de grande taille (LLM) en permettant une récupération sub-linéaire grâce à des techniques d'indexation spécialisées. Ce système répond aux limitations des cadres de mémoire existants qui s'appuient sur des méthodes de récupération exhaustive, pouvant entraîner des problèmes de latence significatifs à mesure que le stockage de la mémoire s'étend.

SwiftMem has been introduced as a query-aware agentic memory system designed to enhance the efficiency of large language model (LLM) agents by enabling sub-linear retrieval through specialized indexing techniques. This system addresses the limitations of existing memory frameworks that rely on exhaustive retrieval methods, which can lead to significant latency issues as memory storage expands.

DYCP: Dynamic Context Pruning for Long-Form Dialogue with LLMs

Was this article worth reading? Share it

LucidQuery AI

ConsoleX

Langfuse

Langtail

Polidict

DMpro

Ready to build your own newsroom?