Track the frontier, from condensed matter to cosmology.
Search papers, labs, and topics across Lattice.
Context Language Models (CLMs) reframe context management by granting models native, unrestricted read-write access to their own context window treated as an editable file. This intrinsic formulation eliminates brittle external agent harnesses, enabling models to optimize their own memory parametrically via online RL—boosting Qwen3.5-9B accuracy on BrowseComp-Plus by 47.6% with 12% fewer FLOPs. Across long-horizon evaluations up to 24 hours, CLMs consistently outperform standard context strategies by up to 65% while reducing serving compute by 35% through custom suffix cache reuse.
Language models that edit their own context window like a mutable file beat external agent scaffolding across long-horizon tasks while cutting compute by up to 59%.
We introduce Context Language Models (CLMs), language models that natively manage their own context. We implement this by treating the context as a file and allowing the model to make unrestricted updates to this file. This allows the model to learn what is most important to maintain in context, and naturally extends to multi-agent systems where multiple agent contexts coexist as files. Building CLMs zero-shot with existing models outperforms SOTA context management strategies across a variety of tasks: 11.4% higher accuracy with 21.5% fewer FLOPs on BrowseComp-Plus, 5% higher scores with 59% fewer FLOPs on 12-hour EdgeBench, and 65% greater improvement with the same compute on a 24-hour multi-repository agent-swarm task. Moreover, by shifting context management from external harness control to intrinsic model behavior, CLMs naturally enable both in-context and parametric learning of context-management strategies. We show that CLMs can be steered with natural-language instructions evolved through a standard skill-optimization loop, improving held-out accuracy by up to 35.9 points on a context-management task while reducing compute. We also introduce an online reinforcement learning method for CLMs, improving Qwen3.5-9B performance on BrowseComp-Plus by 47.6% while using 12% fewer FLOPs. Finally, we co-design Suffix Cache Reuse for CLM serving, further reducing server-side compute by 35% relative to standard SGLang at matched performance.
Meta AI