{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/reinforcement-learning-with-a-corrupted","title":"Reinforcement Learning with a Corrupted Reward Channel","arxiv_id":"1705.08417","date":"2017-05-23","proceeding":null,"authors":["Tom Everitt","Victoria Krakovna","Laurent Orseau","Marcus Hutter","Shane Legg"],"abstract":"No real-world reward function is perfect. Sensory errors and software bugs\nmay result in RL agents observing higher (or lower) rewards than they should.\nFor example, a reinforcement learning agent may prefer states where a sensory\nerror gives it the maximum reward, but where the true reward is actually small.\nWe formalise this problem as a generalised Markov Decision Problem called\nCorrupt Reward MDP. Traditional RL methods fare poorly in CRMDPs, even under\nstrong simplifying assumptions and when trying to compensate for the possibly\ncorrupt rewards. Two ways around the problem are investigated. First, by giving\nthe agent richer data, such as in inverse reinforcement learning and\nsemi-supervised reinforcement learning, reward corruption stemming from\nsystematic sensory errors may sometimes be completely managed. Second, by using\nrandomisation to blunt the agent's optimisation, reward corruption can be\npartially managed under some assumptions.","url_abs":"http://arxiv.org/abs/1705.08417v2","url_pdf":"http://arxiv.org/pdf/1705.08417v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"reinforcement-learning-with-a-corrupted","repo_url":"https://github.com/jvmancuso/safe-grid-agents","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"},{"task_slug":"reinforcement-learning-1","task_name":"Reinforcement Learning (RL)"},{"task_slug":"reinforcement-learning-2","task_name":"reinforcement-learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1705.08417","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}