{"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/arxiv-2510-04019","title":"Simple Policy Gradients for Reasoning with Diffusion Language Models","arxiv_id":"2510.04019","date":"2025-10-05","proceeding":"ICML","authors":["Anthony Zhan"],"abstract":"Diffusion large language models (dLLMs) represent a promising alternative to autoregressive LLMs; however, the lack of effective post-training techniques, including reinforcement learning (RL), remains a key challenge for dLLMs, especially for downstream applications. Existing approaches often rely on a sequence-level view that requires biased likelihood approximations. In this work, we propose Amortized Group Relative Policy Optimization (AGRPO), a policy gradient algorithm that leverages the Markovian nature of dLLMs, optimizing individual denoising steps rather than full sequences. Our approach improves alignment between the trained policy and the inference process and also admits efficient, unbiased gradient updates via a novel timestep estimation scheme. We demonstrate AGRPO's effectiveness on different math and reasoning tasks, achieving absolute accuracy gains of +59.4\\% and +69.7\\% on Countdown and Sudoku over the base LLaDA model, exceeding comparable methods such as diffu-GRPO. Code is available at https://github.com/probablyabot/agrpo.","url_abs":"https://arxiv.org/abs/2510.04019","url_pdf":"https://arxiv.org/pdf/2510.04019","source":{"archive":null,"snapshot":"2025-07-28","note":"not in the Papers with Code archive (frozen at the snapshot)","row_kind":"graph","title_abstract_authors_date":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)"},"code_links":[],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2510.04019","atlas_url":"https://app.syntology.ai/?focus=2510.04019","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2510.04019"}},"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":[{"mentioned_in_github":null,"is_official":null,"provenance":"deterministic:regex_extraction","mentioned_in_paper":null,"url":"https://github.com/probablyabot/agrpo","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"unverified":5},"by_repo_kind":{"found_in_text":{"samples":5,"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":"8c5e4dcd27a0e0d4","entry":"answer_format_reward","repo":"probablyabot/agrpo","repo_kind":"found_in_text","path":"rewards.py","file_url":"https://github.com/probablyabot/agrpo/blob/HEAD/rewards.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":"8c5e4dcd27a0e0d4"}},{"code_sha256_prefix":"4eb074484f3ce4a2","entry":"extract_boxed","repo":"probablyabot/agrpo","repo_kind":"found_in_text","path":"rewards.py","file_url":"https://github.com/probablyabot/agrpo/blob/HEAD/rewards.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":"4eb074484f3ce4a2"}},{"code_sha256_prefix":"a7c4e85e89c9c067","entry":"gumbel_sample","repo":"probablyabot/agrpo","repo_kind":"found_in_text","path":"generate.py","file_url":"https://github.com/probablyabot/agrpo/blob/HEAD/generate.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":"a7c4e85e89c9c067"}},{"code_sha256_prefix":"e30ee20f00809d83","entry":"register_dataset","repo":"probablyabot/agrpo","repo_kind":"found_in_text","path":"data.py","file_url":"https://github.com/probablyabot/agrpo/blob/HEAD/data.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":"e30ee20f00809d83"}},{"code_sha256_prefix":"a5ef0b84542283a7","entry":"thinking_format_reward","repo":"probablyabot/agrpo","repo_kind":"found_in_text","path":"rewards.py","file_url":"https://github.com/probablyabot/agrpo/blob/HEAD/rewards.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":"a5ef0b84542283a7"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.LG","source":"arxiv_api"},"syntology_extracted_results":null}