{"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/trajectory-balance-with-asynchrony-decoupling","title":"Trajectory Balance with Asynchrony: Decoupling Exploration and Learning for Fast, Scalable LLM Post-Training","arxiv_id":"2503.18929","date":"2025-03-24","proceeding":null,"authors":["Brian R. Bartoldson","Siddarth Venkatraman","James Diffenderfer","Moksh Jain","Tal Ben-Nun","Seanie Lee","Minsu Kim","Johan Obando-Ceron","Yoshua Bengio","Bhavya Kailkhura"],"abstract":"Reinforcement learning (RL) is a critical component of large language model (LLM) post-training. However, existing on-policy algorithms used for post-training are inherently incompatible with the use of experience replay buffers, which can be populated scalably by distributed off-policy actors to enhance exploration as compute increases. We propose efficiently obtaining this benefit of replay buffers via Trajectory Balance with Asynchrony (TBA), a massively scalable LLM RL system. In contrast to existing approaches, TBA uses a larger fraction of compute on search, constantly generating off-policy data for a central replay buffer. A training node simultaneously samples data from this buffer based on reward or recency to update the policy using Trajectory Balance (TB), a diversity-seeking RL objective introduced for GFlowNets. TBA offers three key advantages: (1) decoupled training and search, speeding up training wall-clock time by 4x or more; (2) improved diversity through large-scale off-policy sampling; and (3) scalable search for sparse reward settings. On mathematical reasoning, preference-tuning, and automated red-teaming (diverse and representative post-training tasks), TBA produces speed and performance improvements over strong baselines.","url_abs":"https://arxiv.org/abs/2503.18929v1","url_pdf":"https://arxiv.org/pdf/2503.18929v1.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":"trajectory-balance-with-asynchrony-decoupling","repo_url":"https://github.com/bbartoldson/TBA","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"diversity","task_name":"Diversity"},{"task_slug":"large-language-model","task_name":"Large Language Model"},{"task_slug":"mathematical-reasoning","task_name":"Mathematical Reasoning"},{"task_slug":"red-teaming","task_name":"Red Teaming"},{"task_slug":"reinforcement-learning-1","task_name":"Reinforcement Learning (RL)"}],"methods":[{"method_slug":"experience-replay","method_name":"Experience Replay"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2503.18929","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2503.18929"}},"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":"deterministic:regex_extraction","url":"https://github.com/bbartoldson/TBA","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"unverified":10},"by_repo_kind":{"listed":{"samples":10,"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":0,"samples":[{"code_sha256_prefix":"d0b3a48ae72ccfc4","entry":"build_example_dataloader","repo":"bbartoldson/TBA","repo_kind":"listed","path":"src/dist_data_utils.py","file_url":"https://github.com/bbartoldson/TBA/blob/HEAD/src/dist_data_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"d0b3a48ae72ccfc4"}},{"code_sha256_prefix":"2ed37256b49d8561","entry":"example_prepare_dataset","repo":"bbartoldson/TBA","repo_kind":"listed","path":"src/dist_data_utils.py","file_url":"https://github.com/bbartoldson/TBA/blob/HEAD/src/dist_data_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"2ed37256b49d8561"}},{"code_sha256_prefix":"a26a4bc2f82b9ec2","entry":"extract_prediction","repo":"bbartoldson/TBA","repo_kind":"listed","path":"src/gsm8k_utils.py","file_url":"https://github.com/bbartoldson/TBA/blob/HEAD/src/gsm8k_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"a26a4bc2f82b9ec2"}},{"code_sha256_prefix":"150dea51d807fbaa","entry":"format_and_tokenize","repo":"bbartoldson/TBA","repo_kind":"listed","path":"src/gsm8k_utils.py","file_url":"https://github.com/bbartoldson/TBA/blob/HEAD/src/gsm8k_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"150dea51d807fbaa"}},{"code_sha256_prefix":"d70961db4cf293ba","entry":"ignore_prompt_labels","repo":"bbartoldson/TBA","repo_kind":"listed","path":"src/kl.py","file_url":"https://github.com/bbartoldson/TBA/blob/HEAD/src/kl.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"d70961db4cf293ba"}},{"code_sha256_prefix":"5503c179cc4b29b0","entry":"next_free_port","repo":"bbartoldson/TBA","repo_kind":"listed","path":"src/dist_utils.py","file_url":"https://github.com/bbartoldson/TBA/blob/HEAD/src/dist_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"5503c179cc4b29b0"}},{"code_sha256_prefix":"cd83783cc6325cf1","entry":"parse_number","repo":"bbartoldson/TBA","repo_kind":"listed","path":"src/gsm8k_utils.py","file_url":"https://github.com/bbartoldson/TBA/blob/HEAD/src/gsm8k_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"cd83783cc6325cf1"}},{"code_sha256_prefix":"a2e246bb45f45630","entry":"prepare_dataset","repo":"bbartoldson/TBA","repo_kind":"listed","path":"tba_tldr.py","file_url":"https://github.com/bbartoldson/TBA/blob/HEAD/tba_tldr.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"a2e246bb45f45630"}},{"code_sha256_prefix":"e0030346eb43e3a2","entry":"prepare_deepspeed","repo":"bbartoldson/TBA","repo_kind":"listed","path":"src/utils.py","file_url":"https://github.com/bbartoldson/TBA/blob/HEAD/src/utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"e0030346eb43e3a2"}},{"code_sha256_prefix":"9860770ecc9b4ebe","entry":"split_dataset_indices","repo":"bbartoldson/TBA","repo_kind":"listed","path":"src/dist_data_utils.py","file_url":"https://github.com/bbartoldson/TBA/blob/HEAD/src/dist_data_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"9860770ecc9b4ebe"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}