Computer-use agents are usually improved by strengthening perception: better models for reading a screenshot and choosing where to click. Yet a screenshot is only a lossy rendering of the underlying program state, e.g., the files, application backends, and DOM that hold the task data. Different states can produce the same pixels, while code can inspect and modify that state directly. StateAct is a code-first, multi-agent harness built around this distinction. Its main agent works directly with program state by using code, while a dedicated GUI subagent handles screenshot-and-click interaction on the few subgoals that need it, just 28 of 108 tasks and 1.1% of main-agent steps. The same direct access to program state also supports verification: an independent finish gate double-checks the saved result for structural failures, e.g., output that is missing, unsaved, or written to the wrong path. To stay on track over hundreds of steps, the main agent hands subgoals to fresh subagents, keeping its own context focused. On OSWorld 2.0, StateAct lifts Claude Opus 4.8 from 20.6% to 26.9% on binary success, and from 54.8% to 61.6% on partial success, at ~ 9x lower cost per task than the same model driven by screenshots alone; a code-only variant with no GUI subagent reaches only 45.9% partial, below that screenshot-based baseline's 54.8%. In general, grounding action, verification, and memory in state, what we call state-grounding, shifts the main bottleneck from perception toward reasoning: failures depend more on what the agent thinks than on what it sees.
Yan Yang, Xiangru Jian, Ziyang Luo et al.· 0 citations
Reinforcement learning (RL) has become an effective way to improve the tool-use ability of large language models (LLMs), but most existing RL frameworks stop at the policy update. For every new domain, the user is left with two hard systems problems: standing up an isolated environment for each of hundreds of concurrent trajectories and connecting it to training, and scheduling the rollout so that the GPU stays busy across long, multi-turn episodes that spend much of their time stalled on slow tool calls. We present MCP-Universe RL (MCP-U RL), an open-source framework that takes over both. It uses the Model Context Protocol (MCP) as the interface to the environment, so any tool already exposed as an MCP server plugs into training with no RL-specific integration code. It builds the two missing layers once and reuses them across domains: an environment-orchestration layer that provisions, isolates, and recycles the MCP environments over a pluggable container backend, and a rollout-orchestration layer whose staged pipeline overlaps trajectories to keep the GPU busy while episodes wait on tools. A backend-agnostic training layer then applies the update through an existing RL backend, with veRL and slime integrations. With one configuration, changing only the task specification, we train software-engineering, deep-research, and general tool-use agents on gpt-oss-20b and improve task reward in all three.
Ziyang Luo, Yan Yang, Xiangru Jian et al.· 0 citations
Exploring bacteriophage structural diversity is essential for understanding phage biology and for advancing phage-based therapies. Here, we determine the cryo-electron microscopy structure of Jabs, providing, to our knowledge, the first high-resolution view of a phage infecting the multidrug-resistant human pathogen Mycobacterium abscessus. Although Jabs displays the canonical organization of a siphophage, its virion combines several unusual architectural features. The T=9 icosahedral capsid is assembled from two distinct major capsid proteins, with one forming the hexons and the other the pentons, revealing an unprecedented capsid assembly strategy among icosahedral phages. An extensive network of ∼1,700 disulfide bonds stabilize individual structural components and covalently links the capsid, connector, tail, and adhesion device into a continuous assembly. At the distal end of the tail, an elaborate and conformationally dynamic adhesion device comprises multiple candidate receptor-binding proteins organized into complex multidomain architectures, including carbohydrate-binding modules and β-sandwich hetero- and homotrimers resembling the receptor-binding proteins of phages infecting lactic acid bacteria. Together, these findings expand our understanding of phage structural diversity and provide a framework for investigating phage-host interactions and guiding the engineering of therapeutic phages.
Christian Cambillau, Jun Hao Liew, Bernice Siu Yan Tan et al.· bioRxiv· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
