Salesforce's JitMem curates agent memory at read time, gains up to 16.3 points on benchmarks
Salesforce · hf · 2026-09-24
Salesforce introduces Just-in-Time Memory (JitMem), rethinking how LLM agents manage memory. Most existing systems curate memory at write time—distilling finished trajectories into fixed artifacts like reflections or workflows—forcing irreversible storage decisions before the future query is known.
JitMem instead keeps raw trajectories and defers curation to read time: given retrieved traces and the current task, a curator synthesizes a compact, task-adaptive payload. Because the payload is consumed on the same task, the curator can be trained directly from immediate task success, sidestepping long-horizon credit assignment.
Across ALFWorld, WebShop, and τ²-bench, JitMem beats the strongest write-time baselines by 16.2, 16.3, and 3.9 absolute success-rate points. Notably, even an untrained curator matches or surpasses baselines, showing task-adaptive read-time curation is the main source of gains.
More from coding & agent
- Resend joins Stripe Projects: one CLI command wires email into AI agent workflows — jeff_weinstein · 2026-09-24
- MiniMax Code tested: building and deploying a full site from a single prompt — nikola_mr64990 · 2026-09-24
- Agent Reliability QA: The Next AI Category Is Stress-Testing Agents, Not Building Them — DevWithTea123 · 2026-09-24
- Privacy Middleware Concept: Anonymize Sensitive Data Locally Before It Reaches LLMs — Ai_MOON_SHOT · 2026-09-24
- No-BS guide to agent harness engineering: context, guardrails, and feedback — Pavan_Belagatti · 2026-09-24
- Claude Code Projects adds local support: threads can now run on your own machine — bcherny · 2026-09-24