{"url":"/sota/atari-games-on-atari-57","task":{"name":"Atari Games","url":"/task/atari-games","note":null},"dataset":{"name":"Atari-57","url":"/dataset/arcade-learning-environment"},"category":"Playing Games","categories":["Playing Games"],"category_note":null,"description":"The Atari 2600 Games task (and dataset) involves training an agent to achieve high game scores.\r\n\r\n<span style=\"color:grey; opacity: 0.6\">( Image credit: [Playing Atari with Deep Reinforcement Learning](https://arxiv.org/pdf/1312.5602v1.pdf) )</span>","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["Mean Human Normalized Score","Human World Record Breakthrough"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Mean Human Normalized Score":"higher","Human World Record Breakthrough":null}},"counts":{"rows":11,"rows_with_code":5,"rows_with_paper_page":11,"rows_dated":11,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"LBC","metrics":{"Human World Record Breakthrough":"24","Mean Human Normalized Score":"10077.52%"},"uses_additional_data":false,"paper_date":"2023-05-09","paper":"/paper/learnable-behavior-control-breaking-atari","paper_url":"https://arxiv.org/abs/2305.05239v1","paper_title":"Learnable Behavior Control: Breaking Atari Human World Records via Sample-Efficient Behavior Selection","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":2,"model":"GDI-H3(200M frames)","metrics":{"Human World Record Breakthrough":"22","Mean Human Normalized Score":"9620.98%"},"uses_additional_data":false,"paper_date":"2021-06-11","paper":"/paper/gdi-rethinking-what-makes-reinforcement","paper_url":"https://arxiv.org/abs/2106.06232v6","paper_title":"GDI: Rethinking What Makes Reinforcement Learning Different From Supervised Learning","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":3,"model":"GDI-H3","metrics":{"Human World Record Breakthrough":"22","Mean Human Normalized Score":"9620.33%"},"uses_additional_data":false,"paper_date":"2022-06-07","paper":"/paper/generalized-data-distribution-iteration","paper_url":"https://arxiv.org/abs/2206.03192v4","paper_title":"Generalized Data Distribution Iteration","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":4,"model":"GDI-I3","metrics":{"Human World Record Breakthrough":"17","Mean Human Normalized Score":"7810.1%"},"uses_additional_data":false,"paper_date":"2022-06-07","paper":"/paper/generalized-data-distribution-iteration","paper_url":"https://arxiv.org/abs/2206.03192v4","paper_title":"Generalized Data Distribution Iteration","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":5,"model":"MuZero","metrics":{"Human World Record Breakthrough":"19","Mean Human Normalized Score":"4996.20%"},"uses_additional_data":false,"paper_date":"2019-11-19","paper":"/paper/mastering-atari-go-chess-and-shogi-by","paper_url":"https://arxiv.org/abs/1911.08265v2","paper_title":"Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model","code":"https://github.com/werner-duvaud/muzero-general","n_code_links":18,"syntology":{"n_ran":43,"n_unverified":21,"n_samples":64,"n_pointer_only_licence":62}},{"rank_in_archive_order":6,"model":"R2D2","metrics":{"Human World Record Breakthrough":"15","Mean Human Normalized Score":"3374.31%"},"uses_additional_data":false,"paper_date":"2019-05-01","paper":"/paper/recurrent-experience-replay-in-distributed","paper_url":"https://openreview.net/forum?id=r1lyTjAqYX","paper_title":"Recurrent Experience Replay in Distributed Reinforcement Learning","code":"https://github.com/opendilab/DI-engine/blob/main/ding/policy/r2d2.py","n_code_links":3,"syntology":null},{"rank_in_archive_order":7,"model":"LASER","metrics":{"Human World Record Breakthrough":"7","Mean Human Normalized Score":"1741.36%"},"uses_additional_data":false,"paper_date":"2019-09-25","paper":"/paper/off-policy-actor-critic-with-shared-1","paper_url":"https://arxiv.org/abs/1909.11583v2","paper_title":"Off-Policy Actor-Critic with Shared Experience Replay","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":8,"model":"IMPALA, deep","metrics":{"Human World Record Breakthrough":"3","Mean Human Normalized Score":"957.34%"},"uses_additional_data":false,"paper_date":"2018-02-05","paper":"/paper/impala-scalable-distributed-deep-rl-with","paper_url":"http://arxiv.org/abs/1802.01561v3","paper_title":"IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures","code":"https://github.com/ray-project/ray/tree/master/rllib","n_code_links":24,"syntology":{"n_ran":16,"n_unverified":18,"n_samples":34,"n_pointer_only_licence":3}},{"rank_in_archive_order":9,"model":"Rainbow DQN","metrics":{"Human World Record Breakthrough":"4","Mean Human Normalized Score":"873.97%"},"uses_additional_data":false,"paper_date":"2017-10-06","paper":"/paper/rainbow-combining-improvements-in-deep","paper_url":"http://arxiv.org/abs/1710.02298v1","paper_title":"Rainbow: Combining Improvements in Deep Reinforcement Learning","code":"https://github.com/thu-ml/tianshou","n_code_links":34,"syntology":{"n_ran":2,"n_unverified":4,"n_samples":6,"n_pointer_only_licence":1}},{"rank_in_archive_order":10,"model":"M-IQN","metrics":{"Mean Human Normalized Score":"504%"},"uses_additional_data":false,"paper_date":"2020-07-28","paper":"/paper/munchausen-reinforcement-learning","paper_url":"https://arxiv.org/abs/2007.14430v3","paper_title":"Munchausen Reinforcement Learning","code":"https://github.com/google-research/google-research/tree/master/munchausen_rl","n_code_links":6,"syntology":{"n_ran":12,"n_unverified":10,"n_samples":22,"n_pointer_only_licence":0}},{"rank_in_archive_order":11,"model":"GDI-H3","metrics":{},"uses_additional_data":false,"paper_date":"2021-06-11","paper":"/paper/gdi-rethinking-what-makes-reinforcement","paper_url":"https://arxiv.org/abs/2106.06232v6","paper_title":"GDI: Rethinking What Makes Reinforcement Learning Different From Supervised Learning","code":null,"n_code_links":0,"syntology":null}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,264 of the 9,581 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9581,"papers_checked":6264,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":3316},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-24T18:15:14+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":4,"rows_with_any_sample_ran":4,"distinct_papers_with_graph_line":4,"distinct_papers_with_any_sample_ran":4,"samples_over_distinct_papers":{"n_ran":73,"n_unverified":53,"n_samples":126,"n_pointer_only_licence":66,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":73,"n_unverified":53,"n_samples":126,"n_pointer_only_licence":66,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}