AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

Efficient Agent Training for Computer Use

Yanheng He, Jiahe Jin, Pengfei Liu

ICLR 2026 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Human-AI coordination agents · 17 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
augment human trajectories with 9 Claude-synthesized alternative actions per step (Trajectory Boost) instead of human data only (ablation)PC Agent-E (Qwen2.5-VL-72B) · WindowsAgentArena-V2success rate (%): 17.2 -> 36.0Improved
use Trajectory Boost instead of direct end-to-end distillation from Claude 3.7 Sonnet (ablation)PC Agent-E · WindowsAgentArena-V2success rate (%): 26.2 -> 36.0Improved
add 312 human trajectories to direct distillation data (ablation)Direct distillation · WindowsAgentArena-V2success rate (%): 26.2 -> 26.9No clear effect
offline single-step synthesis rather than online trajectory collection for data generationData synthesis · collecting 3120 trajectoriescollection time: 900 hours -> 3 hoursImproved
fine-tune Qwen2.5-VL-72B with PC Agent-E dataQwen2.5-VL-72B · WindowsAgentArena-V2success rate (%): 14.9 -> 36.0Improved
fine-tune the smaller Qwen2.5-VL-7B with the same dataQwen2.5-VL-7B · WindowsAgentArena-V2success rate (%): 5.0 -> 6.4Improved
raise step limit from 15 to 30 (ablation)PC Agent-E · WindowsAgentArena-V2success rate (%): 26.9 -> 36.0Improved
raise step limit from 30 to 50 (ablation)PC Agent-E · WindowsAgentArena-V2success rate (%): 36.0 -> 31.4Hurt
train on Windows data only, evaluate on Linux OSWorldPC Agent-E · OSWorld feasible taskssuccess rate (%): 4.4 -> 10.9Improved
train PC Agent-E on infeasible tasks outcomesPC Agent-E · OSWorld infeasible taskssuccess rate (%): 86.7 -> 63.3Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy