{"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/supervising-strong-learners-by-amplifying","title":"Supervising strong learners by amplifying weak experts","arxiv_id":"1810.08575","date":"2018-10-19","proceeding":null,"authors":["Paul Christiano","Buck Shlegeris","Dario Amodei"],"abstract":"Many real world learning tasks involve complex or hard-to-specify objectives,\nand using an easier-to-specify proxy can lead to poor performance or misaligned\nbehavior. One solution is to have humans provide a training signal by\ndemonstrating or judging performance, but this approach fails if the task is\ntoo complicated for a human to directly evaluate. We propose Iterated\nAmplification, an alternative training strategy which progressively builds up a\ntraining signal for difficult problems by combining solutions to easier\nsubproblems. Iterated Amplification is closely related to Expert Iteration\n(Anthony et al., 2017; Silver et al., 2017), except that it uses no external\nreward function. We present results in algorithmic environments, showing that\nIterated Amplification can efficiently learn complex behaviors.","url_abs":"http://arxiv.org/abs/1810.08575v1","url_pdf":"http://arxiv.org/pdf/1810.08575v1.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":"supervising-strong-learners-by-amplifying","repo_url":"https://github.com/rgreenblatt/ai_alignment_readings","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"supervising-strong-learners-by-amplifying","repo_url":"https://github.com/paulfchristiano/amplification","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null},{"paper_slug":"supervising-strong-learners-by-amplifying","repo_url":"https://github.com/rmoehn/amplification","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1810.08575","atlas_url":"https://app.syntology.ai/?focus=1810.08575","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1810.08575"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/rmoehn/amplification","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/paulfchristiano/amplification","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/rgreenblatt/ai_alignment_readings","reach":{"status":"ok"}}],"summary":{"ran_draft_wrong":3,"unverified":15},"by_repo_kind":{"listed":{"samples":18,"ran":3,"repositories":2}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":15,"samples":[{"code_sha256_prefix":"66c8330623902004","entry":"bind","repo":"paulfchristiano/amplification","repo_kind":"listed","path":"experiment.py","file_url":"https://github.com/paulfchristiano/amplification/blob/HEAD/experiment.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"66c8330623902004"}},{"code_sha256_prefix":"ad5aed415e724ebd","entry":"label","repo":"paulfchristiano/amplification","repo_kind":"listed","path":"experiment.py","file_url":"https://github.com/paulfchristiano/amplification/blob/HEAD/experiment.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"ad5aed415e724ebd"}},{"code_sha256_prefix":"32a0883f9eed30c0","entry":"options","repo":"paulfchristiano/amplification","repo_kind":"listed","path":"experiment.py","file_url":"https://github.com/paulfchristiano/amplification/blob/HEAD/experiment.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"32a0883f9eed30c0"}},{"code_sha256_prefix":"1fea448f5339d7f6","entry":"add_logs","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/tasks/graph.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/tasks/graph.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"1fea448f5339d7f6"}},{"code_sha256_prefix":"ab64ff670a8d3fc4","entry":"dfs","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/tasks/equals.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/tasks/equals.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"ab64ff670a8d3fc4"}},{"code_sha256_prefix":"78df87b7163affcc","entry":"flatten_first","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/models/core.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/models/core.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"78df87b7163affcc"}},{"code_sha256_prefix":"582392fe7ea8743e","entry":"flatten_to","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/models/core.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/models/core.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"582392fe7ea8743e"}},{"code_sha256_prefix":"dff55d08c6c73aca","entry":"get_variable","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/tf_utils.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/tf_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"dff55d08c6c73aca"}},{"code_sha256_prefix":"4c6f302cb61b391d","entry":"ints","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/models/asker.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/models/asker.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"4c6f302cb61b391d"}},{"code_sha256_prefix":"3391fa2577a41f9b","entry":"ints_like","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/models/asker.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/models/asker.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"3391fa2577a41f9b"}},{"code_sha256_prefix":"3d51a440b6d573fd","entry":"is_context","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/tf_utils.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/tf_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"3d51a440b6d573fd"}},{"code_sha256_prefix":"d92ea87a218b4312","entry":"is_known","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/tasks/core.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/tasks/core.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"d92ea87a218b4312"}},{"code_sha256_prefix":"d8d0370822941683","entry":"log","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/tasks/graph.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/tasks/graph.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"d8d0370822941683"}},{"code_sha256_prefix":"3c22c9ecd14f3694","entry":"make_transcript","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/models/asker.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/models/asker.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"3c22c9ecd14f3694"}},{"code_sha256_prefix":"cb49c8ca24a0f6f7","entry":"multinomial","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/models/core.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/models/core.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"cb49c8ca24a0f6f7"}},{"code_sha256_prefix":"4639fc0079859153","entry":"sample","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/tasks/core.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/tasks/core.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"4639fc0079859153"}},{"code_sha256_prefix":"18d8bdd503c65d78","entry":"to_tags","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/tf_utils.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/tf_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"18d8bdd503c65d78"}},{"code_sha256_prefix":"c6ae75112108e1dd","entry":"uniform","repo":"rmoehn/amplification","repo_kind":"listed","path":"amplification/tasks/core.py","file_url":"https://github.com/rmoehn/amplification/blob/HEAD/amplification/tasks/core.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"mcp_get_code":{"code_sha256":"c6ae75112108e1dd"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}