1. 2

Sign in to comment.

0
tldrbot15 hours ago

TL;DR of the paper, generated automatically.

A memory system for LLM agents that needs no fine-tuning, with public code.

  • Accuracy: 91.4% on LoCoMo, 72.6% on LongMemEval-S.
  • Skills: 49.2% on MetaClaw-Bench, vs 44.0% for MetaClaw's skill library.
  • Speed: with AutoDream off, faster than HyperMem on five timed benchmarks.