{"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/self-training-meets-consistency-improving","title":"Self-Training Meets Consistency: Improving LLMs' Reasoning With Consistency-Driven Rationale Evaluation","arxiv_id":"2411.06387","date":"2024-11-10","proceeding":null,"authors":["JaeHyeok Lee","Keisuke Sakaguchi","JinYeong Bak"],"abstract":"Self-training approach for large language models (LLMs) improves reasoning abilities by training the models on their self-generated rationales. Previous approaches have labeled rationales that produce correct answers for a given question as appropriate for training. However, a single measure risks misjudging rationale quality, leading the models to learn flawed reasoning patterns. To address this issue, we propose CREST (Consistency-driven Rationale Evaluation for Self-Training), a self-training framework that further evaluates each rationale through follow-up questions and leverages this evaluation to guide its training. Specifically, we introduce two methods: (1) filtering out rationales that frequently result in incorrect answers on follow-up questions and (2) preference learning based on mixed preferences from rationale evaluation results of both original and follow-up questions. Experiments on three question-answering datasets using open LLMs show that CREST not only improves the logical robustness and correctness of rationales but also improves reasoning abilities compared to previous self-training approaches.","url_abs":"https://arxiv.org/abs/2411.06387v3","url_pdf":"https://arxiv.org/pdf/2411.06387v3.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":"self-training-meets-consistency-improving","repo_url":"https://github.com/jaehyeoklee-119/crest","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"CC-BY-SA-4.0"}}],"tasks":[{"task_slug":"question-answering","task_name":"Question Answering"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2411.06387","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2411.06387"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+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/jaehyeoklee-119/crest","reach":{"status":"ok","spdx":"CC-BY-SA-4.0"}}],"summary":{"unverified":11},"by_repo_kind":{"official":{"samples":11,"ran":0,"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":11,"samples":[{"code_sha256_prefix":"1fa73fef7d214696","entry":"construct_fewshot","repo":"jaehyeoklee-119/crest","repo_kind":"official","path":"src/1_rationale_generation/utils/utils.py","file_url":"https://github.com/jaehyeoklee-119/crest/blob/HEAD/src/1_rationale_generation/utils/utils.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"CC-BY-SA-4.0","inline_ok":false,"mcp_get_code":{"code_sha256":"1fa73fef7d214696"}},{"code_sha256_prefix":"dda1cdac686ed40b","entry":"dataset_collate_fn","repo":"jaehyeoklee-119/crest","repo_kind":"official","path":"src/3_preference_learning/generate_pair_data.py","file_url":"https://github.com/jaehyeoklee-119/crest/blob/HEAD/src/3_preference_learning/generate_pair_data.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"CC-BY-SA-4.0","inline_ok":false,"mcp_get_code":{"code_sha256":"dda1cdac686ed40b"}},{"code_sha256_prefix":"cca9a810e3ed7db1","entry":"eliminate_answer_in_rationale","repo":"jaehyeoklee-119/crest","repo_kind":"official","path":"src/1_rationale_generation/utils/data_processing.py","file_url":"https://github.com/jaehyeoklee-119/crest/blob/HEAD/src/1_rationale_generation/utils/data_processing.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"CC-BY-SA-4.0","inline_ok":false,"mcp_get_code":{"code_sha256":"cca9a810e3ed7db1"}},{"code_sha256_prefix":"9fe518d6e6aa9973","entry":"generate_option_string","repo":"jaehyeoklee-119/crest","repo_kind":"official","path":"src/1_rationale_generation/utils/data_processing.py","file_url":"https://github.com/jaehyeoklee-119/crest/blob/HEAD/src/1_rationale_generation/utils/data_processing.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"CC-BY-SA-4.0","inline_ok":false,"mcp_get_code":{"code_sha256":"9fe518d6e6aa9973"}},{"code_sha256_prefix":"53a6b9163a7b1d29","entry":"generate_option_string","repo":"jaehyeoklee-119/crest","repo_kind":"official","path":"src/1_rationale_generation/utils/utils.py","file_url":"https://github.com/jaehyeoklee-119/crest/blob/HEAD/src/1_rationale_generation/utils/utils.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"CC-BY-SA-4.0","inline_ok":false,"mcp_get_code":{"code_sha256":"53a6b9163a7b1d29"}},{"code_sha256_prefix":"05f62db025293dbc","entry":"generate_prompt_train","repo":"jaehyeoklee-119/crest","repo_kind":"official","path":"src/1_rationale_generation/utils/utils.py","file_url":"https://github.com/jaehyeoklee-119/crest/blob/HEAD/src/1_rationale_generation/utils/utils.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"CC-BY-SA-4.0","inline_ok":false,"mcp_get_code":{"code_sha256":"05f62db025293dbc"}},{"code_sha256_prefix":"d4f06bf817e9b130","entry":"get_paired_dataset","repo":"jaehyeoklee-119/crest","repo_kind":"official","path":"src/3_preference_learning/dpo_model_test_label.py","file_url":"https://github.com/jaehyeoklee-119/crest/blob/HEAD/src/3_preference_learning/dpo_model_test_label.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"CC-BY-SA-4.0","inline_ok":false,"mcp_get_code":{"code_sha256":"d4f06bf817e9b130"}},{"code_sha256_prefix":"0919caaa1dde1eca","entry":"get_paired_dataset","repo":"jaehyeoklee-119/crest","repo_kind":"official","path":"src/3_preference_learning/dpo_training_ratio.py","file_url":"https://github.com/jaehyeoklee-119/crest/blob/HEAD/src/3_preference_learning/dpo_training_ratio.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"CC-BY-SA-4.0","inline_ok":false,"mcp_get_code":{"code_sha256":"0919caaa1dde1eca"}},{"code_sha256_prefix":"1b792ab2a4582d52","entry":"make_output_j","repo":"jaehyeoklee-119/crest","repo_kind":"official","path":"src/3_preference_learning/generate_pair_data.py","file_url":"https://github.com/jaehyeoklee-119/crest/blob/HEAD/src/3_preference_learning/generate_pair_data.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"CC-BY-SA-4.0","inline_ok":false,"mcp_get_code":{"code_sha256":"1b792ab2a4582d52"}},{"code_sha256_prefix":"f3f978ffcca7451b","entry":"make_output_k","repo":"jaehyeoklee-119/crest","repo_kind":"official","path":"src/3_preference_learning/generate_pair_data.py","file_url":"https://github.com/jaehyeoklee-119/crest/blob/HEAD/src/3_preference_learning/generate_pair_data.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"CC-BY-SA-4.0","inline_ok":false,"mcp_get_code":{"code_sha256":"f3f978ffcca7451b"}},{"code_sha256_prefix":"d5a0cc57ae7e3307","entry":"rationale_refining","repo":"jaehyeoklee-119/crest","repo_kind":"official","path":"src/1_rationale_generation/utils/data_processing.py","file_url":"https://github.com/jaehyeoklee-119/crest/blob/HEAD/src/1_rationale_generation/utils/data_processing.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"CC-BY-SA-4.0","inline_ok":false,"mcp_get_code":{"code_sha256":"d5a0cc57ae7e3307"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}