{"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/language-models-learn-to-mislead-humans-via","title":"Language Models Learn to Mislead Humans via RLHF","arxiv_id":"2409.12822","date":"2024-09-19","proceeding":null,"authors":["Jiaxin Wen","Ruiqi Zhong","Akbir Khan","Ethan Perez","Jacob Steinhardt","Minlie Huang","Samuel R. Bowman","He He","Shi Feng"],"abstract":"Language models (LMs) can produce errors that are hard to detect for humans, especially when the task is complex. RLHF, the most popular post-training method, may exacerbate this problem: to achieve higher rewards, LMs might get better at convincing humans that they are right even when they are wrong. We study this phenomenon under a standard RLHF pipeline, calling it \"U-SOPHISTRY\" since it is Unintended by model developers. Specifically, we ask time-constrained (e.g., 3-10 minutes) human subjects to evaluate the correctness of model outputs and calculate humans' accuracy against gold labels. On a question-answering task (QuALITY) and programming task (APPS), RLHF makes LMs better at convincing our subjects but not at completing the task correctly. RLHF also makes the model harder to evaluate: our subjects' false positive rate increases by 24.1% on QuALITY and 18.3% on APPS. Finally, we show that probing, a state-of-the-art approach for detecting Intended Sophistry (e.g. backdoored LMs), does not generalize to U-SOPHISTRY. Our results highlight an important failure mode of RLHF and call for more research in assisting humans to align them.","url_abs":"https://arxiv.org/abs/2409.12822v3","url_pdf":"https://arxiv.org/pdf/2409.12822v3.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":"language-models-learn-to-mislead-humans-via","repo_url":"https://github.com/jiaxin-wen/misleadlm","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"question-answering","task_name":"Question Answering"}],"methods":[{"method_slug":"align","method_name":"ALIGN"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2409.12822","atlas_url":"https://app.syntology.ai/?focus=2409.12822","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2409.12822"}},"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/jiaxin-wen/misleadlm","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran":4,"unverified":2},"by_repo_kind":{"official":{"samples":6,"ran":4,"repositories":1}},"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":0,"samples":[{"code_sha256_prefix":"2ab02dee00149590","entry":"make_head","repo":"jiaxin-wen/misleadlm","repo_kind":"official","path":"trlx/utils/modeling.py","file_url":"https://github.com/jiaxin-wen/misleadlm/blob/HEAD/trlx/utils/modeling.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"2ab02dee00149590"}},{"code_sha256_prefix":"e2c3cae9e92c72e8","entry":"parse_result","repo":"jiaxin-wen/misleadlm","repo_kind":"official","path":"trlx/ray_tune/wandb.py","file_url":"https://github.com/jiaxin-wen/misleadlm/blob/HEAD/trlx/ray_tune/wandb.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"e2c3cae9e92c72e8"}},{"code_sha256_prefix":"ace162af010af5d8","entry":"rhasattr","repo":"jiaxin-wen/misleadlm","repo_kind":"official","path":"trlx/utils/modeling.py","file_url":"https://github.com/jiaxin-wen/misleadlm/blob/HEAD/trlx/utils/modeling.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"ace162af010af5d8"}},{"code_sha256_prefix":"520cf375ca747e7f","entry":"tokenize_dialogue","repo":"jiaxin-wen/misleadlm","repo_kind":"official","path":"trlx/pipeline/offline_pipeline.py","file_url":"https://github.com/jiaxin-wen/misleadlm/blob/HEAD/trlx/pipeline/offline_pipeline.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"520cf375ca747e7f"}},{"code_sha256_prefix":"c80bffc6c3b486fc","entry":"create_report","repo":"jiaxin-wen/misleadlm","repo_kind":"official","path":"trlx/ray_tune/wandb.py","file_url":"https://github.com/jiaxin-wen/misleadlm/blob/HEAD/trlx/ray_tune/wandb.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"c80bffc6c3b486fc"}},{"code_sha256_prefix":"fc2499cfa7162602","entry":"rgetattr","repo":"jiaxin-wen/misleadlm","repo_kind":"official","path":"trlx/utils/modeling.py","file_url":"https://github.com/jiaxin-wen/misleadlm/blob/HEAD/trlx/utils/modeling.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"fc2499cfa7162602"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}