{"about":{"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.","site":"https://codewithpapers.app","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","syntology":{"site":"https://syntology.ai","developers":"https://syntology.ai/developers","mcp":{"server":"https://syntology.ai/mcp","transport":"streamable-http","server_card":"https://syntology.ai/.well-known/mcp/server-card.json","auth":{"type":"trial token, no account","trial_token":"https://syntology.ai/api/oauth/trial/token","method":"POST","docs":"https://syntology.ai/developers"}},"have":"https://syntology.ai/api/graph/have?x=<method, arXiv id or title> (free, answers coverage only)","paper_base":"https://syntology.ai/paper/","atlas_base":"https://app.syntology.ai/?focus="},"machine_readable":[{"url":"https://codewithpapers.app/llms.txt","what":"the machine catalog: every machine-readable file, counted"},{"url":"https://codewithpapers.app/index/manifest.json","what":"paper-to-code index by arXiv id, with Syntology's counts"},{"url":"https://codewithpapers.app/search/manifest.json","what":"site search index (titles, authors) and its files"},{"url":"https://codewithpapers.app/download","what":"bulk files: Syntology's layer, described there"},{"url":"https://codewithpapers.app/build_manifest.json","what":"the build record: inputs, counts, exclusions, probes"}]},"url":"/paper/trial-without-error-towards-safe","title":"Trial without Error: Towards Safe Reinforcement Learning via Human Intervention","arxiv_id":"1707.05173","date":"2017-07-17","proceeding":null,"authors":["William Saunders","Girish Sastry","Andreas Stuhlmueller","Owain Evans"],"abstract":"AI systems are increasingly applied to complex tasks that involve interaction\nwith humans. During training, such systems are potentially dangerous, as they\nhaven't yet learned to avoid actions that could cause serious harm. How can an\nAI system explore and learn without making a single mistake that harms humans\nor otherwise causes serious damage? For model-free reinforcement learning,\nhaving a human \"in the loop\" and ready to intervene is currently the only way\nto prevent all catastrophes. We formalize human intervention for RL and show\nhow to reduce the human labor required by training a supervised learner to\nimitate the human's intervention decisions. We evaluate this scheme on Atari\ngames, with a Deep RL agent being overseen by a human for four hours. When the\nclass of catastrophes is simple, we are able to prevent all catastrophes\nwithout affecting the agent's learning (whereas an RL baseline fails due to\ncatastrophic forgetting). However, this scheme is less successful when\ncatastrophes are more complex: it reduces but does not eliminate catastrophes\nand the supervised learner fails on adversarial examples found by the agent.\nExtrapolating to more challenging environments, we show that our implementation\nwould not scale (due to the infeasible amount of human labor required). We\noutline extensions of the scheme that are necessary if we are to train\nmodel-free agents without a single catastrophe.","url_abs":"http://arxiv.org/abs/1707.05173v1","url_pdf":"http://arxiv.org/pdf/1707.05173v1.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":"trial-without-error-towards-safe","repo_url":"https://github.com/gsastry/human-rl","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"atari-games","task_name":"Atari Games"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"},{"task_slug":"reinforcement-learning-1","task_name":"Reinforcement Learning (RL)"},{"task_slug":"safe-reinforcement-learning","task_name":"Safe Reinforcement Learning"},{"task_slug":"reinforcement-learning-2","task_name":"reinforcement-learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1707.05173","atlas_url":"https://app.syntology.ai/?focus=1707.05173","mcp":null,"developers":"https://syntology.ai/developers","held":true,"reason":"arxiv_id_in_atlas","checked_against":{"atlas_ids":549449,"atlas_ids_sha256":"afd72399dafd9906e9062927f70655952385a3d75488558e1f0484e950ab09b7","read_at":"2026-10-05T01:46:21Z"},"agent_calls":[{"tool":"get_citation_path","arguments":{"paper_1":"1707.05173"},"arguments_in":{"paper_2":"another paper's arXiv id or title"},"call":"get_citation_path(paper_1=\"1707.05173\", paper_2=\"…\")"},{"tool":"get_concepts_for_paper","arguments":{"arxiv_id":"1707.05173"},"call":"get_concepts_for_paper(arxiv_id=\"1707.05173\")"}],"by_repo":[{"repo":"gsastry/human-rl","kind":"official","samples":0,"ran":0,"instrument":0,"unverified":0}],"by_repo_is":"one row per repository in the archive's code links, Syntology's graph links or the samples; samples 0 means no sample is linked to this paper, a statement about Syntology's coverage, not about the repository: with harvested_for_other_papers present, Syntology harvested the repository for that many other papers, and without it, Syntology harvested nothing from it; placed_by_identical_code counts samples whose link records no repository, placed here because identical code was harvested from this repository for another paper; ran counts executions on a synthesized input and instrument counts failures of Syntology's instrument, not of the code"},"code_links_note":null,"arxiv_metadata":null,"syntology_extracted_results":null,"record_sha256":"e36476328d58919bd83a2ba9b2771429744461fab4ed56f119ce0e3f350a808d","record_changed_at":"2026-09-28","record_changed_at_basis":"first_hashed"}