2026-11-13

AI for Software Development

No paper recommendations for this topic.

Harness Engineering

No paper recommendations for this topic.

Human-in-the-Loop Evaluation

No paper recommendations for this topic.

Human-AI Collaboration

Interactive Visual Gym: Evaluating Visual Interaction Models with Simulated Users in Interactive Worlds

Relevance: Targets visual assistants that support users during ongoing physical tasks, where assistant responses shape subsequent user behavior. It emphasizes closed-loop collaborative interaction, including guidance, latency, and recovery from user execution errors. This is a direct study of human-assistant teaming dynamics rather than an isolated autonomous agent.

💡 Summary 📄 Full paper

Simulated Users

Interactive Visual Gym: Evaluating Visual Interaction Models with Simulated Users in Interactive Worlds

Relevance: Its central contribution is a persona-conditioned user simulation harness that collaborates with the assistant and injects controlled execution errors and out-of-plan behaviors. It is validated against human preferences and used to train via GRPO in closed loop. This directly matches the use of simulated users for interactive evaluation and training.

💡 Summary 📄 Full paper