Salesforce research: extracting agent memory just-in-time beats summarizing, +16 success rate
dair_ai · x · 2026-09-26
Salesforce AI Research's new paper on agent memory proposes Just-in-Time Memory: store raw trajectories instead of summarizing each run at its end, and decide what to extract when the next task arrives.
- A curator reads retrieved traces together with the new task and writes a short memory payload for that task
- Since the payload is used immediately, the curator can be trained on whether that same task succeeds
- Beats the strongest baseline by 16.2, 16.3 and 3.9 success-rate points on ALFWorld, WebShop and tau2-bench
- Even an untrained curator matches or beats memory written at task end
More from coding & agent
- Sydney Codex Community Buildathon set for Nov 14 with Cerebras-backed ultrafast compute track — paw_lean · 2026-11-11
- shadcn's improve hits 10k stars: let top models plan, cheap models execute code fixes — shadcn · 2026-11-11
- OpenAI open-sources Codex Security CLI and SDK for AI-driven vulnerability scanning and fixing — openai · 2026-11-11
- Claude Opus 5.5 runs an architecture search and finds AttnRes-Canon, a new top Transformer design — SonglinYang4 · 2026-11-11
- 5 research habits worth knowing: read papers backwards, reproduce first, wire coding agents via MCP — cneuralnetwork · 2026-11-11
- 'Adam can now do CAE': AI agent showcase in engineering simulation draws buzz — burhop · 2026-11-11