ScreenSpot-v2 instruction variants
Instruction variants over ScreenSpot-v2 screenshots that distinguish original commands, explicit actions, target descriptions and impossible operations.
Source-reviewed research and assets tagged with Robustness.
Membership follows reviewed tags; it does not establish a dependency between neighboring records. Dates preserve source precision. Annotations retain their original language.
Instruction variants over ScreenSpot-v2 screenshots that distinguish original commands, explicit actions, target descriptions and impossible operations.
Windows benchmark revision with task/evaluator fixes and VM-initialization support, documented alongside PC-Agent-E.
611 desktop GUI tasks that agents must complete from varied starting points, shipped with instructional videos and project files, plus the self-reflective WorldGUI-Agent baseline.
Announced GUI trajectory synthesis resources for recovery from policy-induced errors; the inspected repository marks code and data releases as in preparation.
Research code for constructing and evaluating adversarial inputs to web agents, with a black-box red-team pipeline using Mind2Web data and SeeAct.
Optimizes GUI-agent execution code from repeated rollout evidence and checks proposed changes against fixed validation gates.
Showing 181–186 of 186 resources.
Get full records and evidence through the topic export manifest. Source records can be passed to the standard RAG and citation exporters after cloning the catalog.