{"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/the-accuracy-paradox-in-rlhf-when-better","title":"The Accuracy Paradox in RLHF: When Better Reward Models Don't Yield Better Language Models","arxiv_id":"2410.06554","date":"2024-10-09","proceeding":null,"authors":["Yanjun Chen","Dawei Zhu","Yirong Sun","Xinghao Chen","Wei zhang","Xiaoyu Shen"],"abstract":"Reinforcement Learning from Human Feedback significantly enhances Natural Language Processing by aligning language models with human expectations. A critical factor in this alignment is the strength of reward models used during training. This study explores whether stronger reward models invariably lead to better language models. In this paper, through experiments on relevance, factuality, and completeness tasks using the QA-FEEDBACK dataset and reward models based on Longformer, we uncover a surprising paradox: language models trained with moderately accurate reward models outperform those guided by highly accurate ones. This challenges the widely held belief that stronger reward models always lead to better language models, and opens up new avenues for future research into the key factors driving model performance and how to choose the most suitable reward models. Code and additional details are available at https://github.com/EIT-NLP/AccuracyParadox-RLHF.","url_abs":"https://arxiv.org/abs/2410.06554v2","url_pdf":"https://arxiv.org/pdf/2410.06554v2.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":"the-accuracy-paradox-in-rlhf-when-better","repo_url":"https://github.com/EIT-NLP/AccuracyParadox-RLHF","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[],"methods":[{"method_slug":"adamw","method_name":"AdamW"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"attention-dropout","method_name":"Attention Dropout"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"linear-warmup-with-linear-decay","method_name":"Linear Warmup With Linear Decay"},{"method_slug":"longformer","method_name":"Longformer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"weight-decay","method_name":"Weight Decay"},{"method_slug":"wordpiece","method_name":"WordPiece"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2410.06554","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.06554"}},"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/EIT-NLP/AccuracyParadox-RLHF","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran":3,"ran_draft_wrong":2,"unverified":4},"by_repo_kind":{"official":{"samples":9,"ran":5,"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":"ce0ed06559b93e81","entry":"create_position_ids_from_input_ids","repo":"EIT-NLP/AccuracyParadox-RLHF","repo_kind":"official","path":"reward_modeling/my_longformer.py","file_url":"https://github.com/EIT-NLP/AccuracyParadox-RLHF/blob/HEAD/reward_modeling/my_longformer.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":"ce0ed06559b93e81"}},{"code_sha256_prefix":"e5bacba887a721ef","entry":"reduce_mean","repo":"EIT-NLP/AccuracyParadox-RLHF","repo_kind":"official","path":"fgrlhf/utils.py","file_url":"https://github.com/EIT-NLP/AccuracyParadox-RLHF/blob/HEAD/fgrlhf/utils.py","link_basis":"plan_row","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"e5bacba887a721ef"}},{"code_sha256_prefix":"174bfbbeb1f29e12","entry":"reduce_std","repo":"EIT-NLP/AccuracyParadox-RLHF","repo_kind":"official","path":"fgrlhf/utils.py","file_url":"https://github.com/EIT-NLP/AccuracyParadox-RLHF/blob/HEAD/fgrlhf/utils.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":"174bfbbeb1f29e12"}},{"code_sha256_prefix":"5874b0a8910897fd","entry":"reduce_sum","repo":"EIT-NLP/AccuracyParadox-RLHF","repo_kind":"official","path":"fgrlhf/utils.py","file_url":"https://github.com/EIT-NLP/AccuracyParadox-RLHF/blob/HEAD/fgrlhf/utils.py","link_basis":"plan_row","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"5874b0a8910897fd"}},{"code_sha256_prefix":"d37f5dbfe7b33d1e","entry":"tril_bfloat16","repo":"EIT-NLP/AccuracyParadox-RLHF","repo_kind":"official","path":"reward_modeling/my_longformer.py","file_url":"https://github.com/EIT-NLP/AccuracyParadox-RLHF/blob/HEAD/reward_modeling/my_longformer.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":"d37f5dbfe7b33d1e"}},{"code_sha256_prefix":"b3257abd722edb60","entry":"get_rouge_scores","repo":"EIT-NLP/AccuracyParadox-RLHF","repo_kind":"official","path":"fgrlhf/evaluators.py","file_url":"https://github.com/EIT-NLP/AccuracyParadox-RLHF/blob/HEAD/fgrlhf/evaluators.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":"b3257abd722edb60"}},{"code_sha256_prefix":"94a55f838c975848","entry":"postprocess_text","repo":"EIT-NLP/AccuracyParadox-RLHF","repo_kind":"official","path":"fgrlhf/evaluators.py","file_url":"https://github.com/EIT-NLP/AccuracyParadox-RLHF/blob/HEAD/fgrlhf/evaluators.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":"94a55f838c975848"}},{"code_sha256_prefix":"e7f95d91543ac720","entry":"split_text_to_sentences","repo":"EIT-NLP/AccuracyParadox-RLHF","repo_kind":"official","path":"fgrlhf/reward_utils.py","file_url":"https://github.com/EIT-NLP/AccuracyParadox-RLHF/blob/HEAD/fgrlhf/reward_utils.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":"e7f95d91543ac720"}},{"code_sha256_prefix":"11589c4018cfc261","entry":"split_text_to_subsentences","repo":"EIT-NLP/AccuracyParadox-RLHF","repo_kind":"official","path":"fgrlhf/reward_utils.py","file_url":"https://github.com/EIT-NLP/AccuracyParadox-RLHF/blob/HEAD/fgrlhf/reward_utils.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":"11589c4018cfc261"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}