Context Language Models paper: LLMs natively manage their own context as editable files
alex_verem · x · 2026-10-05
The arXiv paper (2609.37725) behind the CLM thread, by Rulin Shao, Pang Wei Koh, Luke Zettlemoyer et al.: models treat context as a file and make unrestricted updates, with multi-agent contexts coexisting as files.
Numbers from the abstract: 11.4% higher accuracy and 21.5% fewer FLOPs on BrowseComp-Plus; +5 points with 59% fewer FLOPs on 12-hour EdgeBench; 65% greater improvement on a 24-hour multi-repo agent-swarm task. Skill-optimized natural-language steering adds up to 35.9 held-out points; online RL lifts Qwen3.5-9B by 47.6% with 12% fewer FLOPs. Suffix Cache Reuse cuts server-side compute by 35% vs standard SGLang.
More from coding & agent
- Dev rebuilt his coding agents' UI as iMessage and says life is better now — davidfromkansas · 2026-10-05
- Redditor builds offline AI agent network across Legion laptop and iPhone with Tailscale — Due_Recording_5802 · 2026-10-05
- Every MCP tool passed in isolation, yet the workflow still failed — test full chains — Stock-Pumpkin-8859 · 2026-10-05
- Instead of booking flights, these are the agent demo tasks people actually want — vivekhaldar · 2026-10-05
- Amazon (2B) and Cloudflare (9B/27B) ship open-source decision models rivaling Jev — mark_k · 2026-10-05
- Autonomous agents only pay off where value justifies engineer escalation, dev argues — Pavel_Asparagus · 2026-10-05