{"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/debating-with-more-persuasive-llms-leads-to","title":"Debating with More Persuasive LLMs Leads to More Truthful Answers","arxiv_id":"2402.06782","date":"2024-02-09","proceeding":null,"authors":["Akbir Khan","John Hughes","Dan Valentine","Laura Ruis","Kshitij Sachan","Ansh Radhakrishnan","Edward Grefenstette","Samuel R. Bowman","Tim Rocktäschel","Ethan Perez"],"abstract":"Common methods for aligning large language models (LLMs) with desired behaviour heavily rely on human-labelled data. However, as models grow increasingly sophisticated, they will surpass human expertise, and the role of human evaluation will evolve into non-experts overseeing experts. In anticipation of this, we ask: can weaker models assess the correctness of stronger models? We investigate this question in an analogous setting, where stronger models (experts) possess the necessary information to answer questions and weaker models (non-experts) lack this information. The method we evaluate is debate, where two LLM experts each argue for a different answer, and a non-expert selects the answer. We find that debate consistently helps both non-expert models and humans answer questions, achieving 76% and 88% accuracy respectively (naive baselines obtain 48% and 60%). Furthermore, optimising expert debaters for persuasiveness in an unsupervised manner improves non-expert ability to identify the truth in debates. Our results provide encouraging empirical evidence for the viability of aligning models with debate in the absence of ground truth.","url_abs":"https://arxiv.org/abs/2402.06782v4","url_pdf":"https://arxiv.org/pdf/2402.06782v4.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":"debating-with-more-persuasive-llms-leads-to","repo_url":"https://github.com/ucl-dark/llm_debate","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"persuasiveness","task_name":"Persuasiveness"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2402.06782","atlas_url":"https://app.syntology.ai/?focus=2402.06782","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2402.06782"}},"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/ucl-dark/llm_debate","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran":11,"unverified":2},"by_repo_kind":{"official":{"samples":13,"ran":11,"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":"fb8e607626aa645c","entry":"add_assistant_message","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/llm_api/base_llm.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/llm_api/base_llm.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":"fb8e607626aa645c"}},{"code_sha256_prefix":"c6dac245fac2c7f1","entry":"convert_to_prob","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/llm_api/base_llm.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/llm_api/base_llm.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":"c6dac245fac2c7f1"}},{"code_sha256_prefix":"56346bcde59d7a4a","entry":"count_tokens","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/llm_api/anthropic_llm.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/llm_api/anthropic_llm.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"56346bcde59d7a4a"}},{"code_sha256_prefix":"5114a76c19ed4e1f","entry":"load_secrets","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/utils.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/utils.py","link_basis":"harvester_set","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":"5114a76c19ed4e1f"}},{"code_sha256_prefix":"a055cc7323a5cf50","entry":"load_yaml","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/utils.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/utils.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":"a055cc7323a5cf50"}},{"code_sha256_prefix":"4ae8e4de76e4be99","entry":"load_yaml_cached","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/utils.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/utils.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":"4ae8e4de76e4be99"}},{"code_sha256_prefix":"f9349b8202644fd7","entry":"pair_players","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/swiss_tournament.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/swiss_tournament.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":"f9349b8202644fd7"}},{"code_sha256_prefix":"e176ec2bc15a5a9f","entry":"play_round","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/swiss_tournament.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/swiss_tournament.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":"e176ec2bc15a5a9f"}},{"code_sha256_prefix":"f8bbc47c2b491933","entry":"price_per_token","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/llm_api/anthropic_llm.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/llm_api/anthropic_llm.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"f8bbc47c2b491933"}},{"code_sha256_prefix":"ea74105eb20142a0","entry":"price_per_token","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/llm_api/openai_llm.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/llm_api/openai_llm.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":"ea74105eb20142a0"}},{"code_sha256_prefix":"c0858383ecfb0a10","entry":"swiss_tournament","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/swiss_tournament.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/swiss_tournament.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":"c0858383ecfb0a10"}},{"code_sha256_prefix":"3b94d7f86b74da2b","entry":"count_tokens","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/llm_api/openai_llm.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/llm_api/openai_llm.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":"3b94d7f86b74da2b"}},{"code_sha256_prefix":"e5748ed4cebfd681","entry":"make_exp_dir","repo":"ucl-dark/llm_debate","repo_kind":"official","path":"core/tournament.py","file_url":"https://github.com/ucl-dark/llm_debate/blob/HEAD/core/tournament.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":"e5748ed4cebfd681"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}