Paper: knowledge-graph memory with forgetting beats flat vector retrieval for agents
burkov · x · 2026-09-07
A new paper asks whether representing conversational history as a knowledge graph improves retrieval quality for long-term LLM agents versus flat vector retrieval, and whether an explicit forgetting mechanism can keep the graph from growing without hurting performance. The motivation: agent memory stores balloon over months of interaction, raising costs and noise while flat retrieval struggles with relational and multi-hop questions.
The pipeline extracts typed nodes and attributed edges per dialogue turn, retrieves a two-hop subgraph around the top-5 matching entities, and periodically prunes nodes whose importance score (recency, access frequency, degree, age) falls below a threshold, evaluated on a 500-question benchmark.
More from coding & agent
- Vibe Coding plagiarism drama: 'Never seen it before' isn't originality, dev argues — ezshine · 2026-09-07
- MobileWorld benchmark: 201-task eval of phone GUI agents, best combo only ~52% — East-Muffin-6472 · 2026-09-07
- Redditor builds local agent 'Sai' with a real body, thermal senses and skin in the game — Quebber · 2026-09-07
- An agent workspace that combines cloud desktop, file management and custom agents — tekbog · 2026-09-07
- Making Codex the brain and Gemini the writer: one engineer's path to self-improving skills — xiaohu · 2026-09-07
- GPT-6 Astra ports Pokemon Tower Defense from Flash to Phaser 3 in ~5 hours, 4 prompts — TheReal4982 · 2026-09-07