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
- Genspark launches GenCode coding agent with open-weight models at 1/10 the cost — baseten · 2026-09-26
- Agent UX Is Converging, but "Context Proximity" Still Drives Which Assistant Users Reach For — manosaie · 2026-09-26
- Tiny 4B model drives a delivery car: one decision per junction, no 500-token monologue — MaziyarPanahi · 2026-09-26
- Stop Burning Credits: Why You Should Start Every Lovable Project in Chat Mode — damienghader · 2026-09-26
- skill-audit: Open-source pre-install auditor targets rising agent skill supply-chain risk — masiha97 · 2026-09-26
- SimonW: coding agents make software engineering harder, demanding extraordinary discipline — blaizedsouza · 2026-09-26