OSWorld

Timeline

1. Oct 15, 2026 · Research paper · 2 sources
   LEASH: Uncertainty-Aware Weak-to-Strong Escalation and Oversight for Computer-Use Agents
   To this end, we propose LEASH, a training-free framework that unleashes the strong agent only when the weak agent's latent uncertainty signals it is needed, and keeps it on a leash afterward.
2. Oct 15, 2026 · Open-source release · 2 sources
   OpenGCUA: Toward a Fully Open Stack for Generalist Computer Use
   We introduce Open Generalist CUA (OpenGCUA), an ongoing project to develop and open the full stack for GCUAs with the community.
3. Oct 14, 2026 · Model release · 10 sources
   [AINews] Meta Muse Wonder & Wonder Full: MSL’s Claude 5.5-tier Frontier Model
   We have been coveringe MSL’s 14 month journey to becoming a frontier lab, but only today did the final ball drop: Meta’s Muse Wonder to match Spark and Glimmer, the way that OpenAI has Sol/Terra/Luna and Anthropic has Opus/Sonnet/Haiku.
4. Oct 14, 2026 · Research paper · 1 source
   JointServe: Serving Multi-Model Agents on Memory-Constrained GPUs
   We present JointServe, a serving system that treats weight shards and KV blocks as residency objects competing for one unified memory budget.
5. Oct 13, 2026 · Research paper · 2 sources
   GroundGUI: GUI Grounding by High-Quality Data and Distillation-to-Reinforcement Learning
   To bridge this gap, we introduce the Screen Referring Grounding Dataset (ScreenRef), a comprehensive GUI grounding dataset consisting of GUI screenshots paired with natural language referring expressions and box or point annotations.
6. Oct 9, 2026 · Research paper · 1 source
   ConClick: Element-Wise Contrastive Learning Improves GUI Grounding
   Motivated by these findings, we propose ConClick, a contrastive learning framework that teaches models to distinguish the target element from plausible distractors.
7. Oct 9, 2026 · Research paper · 2 sources
   Work2Skill: Capturing Human GUI Experience for Reusable Agent Skill Induction
   We thus introduce Work2Skill, a framework that captures human GUI experience for reusable agent skill induction.
8. Oct 8, 2026 · Research paper · 2 sources
   SoftwareGym: Scaling Computer-Use Agent Training with Diverse Synthesized Software Environments
   Computer-use agents (CUAs) are rapidly improving, yet synthesizing diverse and reliable software environments at scale remains challenging.
9. Sep 28, 2026 · Model release · 1 source
   Holo4: powering generalist computer-use agents
   Holo4 is our new series of agentic models.
10. Jul 21, 2026 · Product / feature launch · 1 source
    Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
    Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

Often appears with