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
- New TLAPS-Bench blog: formally proving correctness of real-world system specs with TLA+ — tianyin_xu · 2026-11-19
- Pierre releases Words, a streaming Markdown parser 2x faster than md4x and 20x faster than Streamdown — andersonbcdefg · 2026-11-19
- Ex-Claude superfan: getting the best from Astra/Sol requires relearning how you prompt — mariofilhoml · 2026-11-19
- Rivet's lesson: JEV doesn't reason—split severity classification into tiny yes/no prompts — yawnxyz · 2026-11-19
- 16x Microsoft MVP demos a native PowerShell agent built on Microsoft's Intelligent Terminal — dfinke · 2026-11-19
- Musk: Treat Grok Bot like a really smart employee, says most users underuse it — elonmusk · 2026-11-19