Supporting dataset for Visual Bias in Simulated Users: The Impact of Luminance and Contrast on Reinforcement Learning-based Interaction

This dataset contains simulator folders compatible with the User-in-the-Box framework for simulating human-computer-interaction. It includes final training checkpoints of reinforcement learning agents trained with distinct luminance values for task-relevant objects, distractors and background across two interaction tasks, pointing and tracking, with 3 variants each: without distractors, with moving and static distractors. The RL agents were trained within the User-in-the-Box simulation environment. These checkpoints enable researchers to reproduce experiments, analyze learned policies, and investigate the influence of reward function design on agent behavior. To evaluate a checkpoint, copy the simulator folder (starting with mobl_arms_index) into the simulators directory and run the following command: python uitb/test/evaluator.py simulators/mobl_arms_index_pointing --record --logging Note: The evaluation script evaluator.py and the simulators folder is part of the User-in-the-Box repository, which must be cloned beforehand. For detailed information on implementation and experimental procedures, please refer to the associated publication.