LExGraph advances exploration memory
Original title: Twists and Turns: A Graph-Based Benchmark and Framework for Long-Term Memory in Exploration Tasks
LExGraph organizes exploration dialogues into graphs for superior long-term memory.
Exploration tasks form graphs with revisions and branches. LExMem is a graph-guided benchmark with 224 dialogues and 2,018 QA instances. LExGraph incrementally organizes dialogues into memory nodes using hybrid retrieval. It outperforms existing systems and generalizes to external benchmarks.
- 224 dialogues and 2,018 QA instances
- Graph-based evolutionary mechanism
- Incremental memory node organization
- Hybrid retrieval and neighbor expansion
- Outperforms retrieval systems
From the original article
1 min read at arxiv.orgLong-term problem-solving processes, such as scientific research and software development, proceed along an evolving trial-and-error trajectory toward a central goal. We term these \textit{exploration tasks}, which require models to remember not only facts but also how solutions are tested against intermediate outcomes and early attempts are continually revised or revived.
By Chongyuan Dai · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.