{"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-2605-16941","title":"Roll Out and Roll Back: Diffusion LLMs are Their Own Efficiency Teachers","arxiv_id":"2605.16941","date":"2026-05-16","proceeding":null,"authors":["Fanqin Zeng","Feng Hong","Geng Yu","Huangjie Zheng","Xiaofeng Cao","Ya Zhang","Bo Han","Yanfeng Wang","Jiangchao Yao"],"abstract":"Diffusion Large Language Models (DLLMs) promise fast parallel generation, yet open-source DLLMs still face a severe quality-speed trade-off: accelerating decoding by revealing multiple tokens often causes substantial quality degradation. We attribute this dilemma to a train-inference mismatch amplified by irreversible decoding. While training reconstructs tokens from randomly corrupted states, efficient inference requires an adaptive denoising order, where easier tokens are revealed earlier and context-dependent ones are deferred. This view motivates two complementary methods: an inference-time method that makes parallel decoding revokable, and a training-time extension that distills the reliable order exposed by this revokable process. Accordingly, we first propose Wide-In, Narrow-Out (WINO), a training-free decoding algorithm that enables revokable parallel generation. WINO aggressively drafts multiple tokens, verifies generated tokens with enriched global context, and re-masks unreliable ones for later refinement. Building on this discovered order, we further introduce WINO+, which injects the verified denoising trajectories produced by WINO into model parameters, aligning training with efficient inference. Experiments on LLaDA and MMaDA show that WINO improves both quality and efficiency, while WINO+ further strengthens this progression. On GSM8K, WINO improves accuracy from 73.24% to 75.82% with a 6.10x step reduction, and WINO+ further achieves 76.58% with a 6.83x reduction. On Flickr30K, WINO+ reaches a 16.22x step reduction with improved CIDEr. These results demonstrate that DLLMs can serve as their own efficiency teachers by first discovering reliable denoising orders through revokable decoding and then learning to follow them for faster generation. Code is available at https://github.com/Feng-Hong/WINO-DLLM/tree/WINO-plus.","url_abs":"https://arxiv.org/abs/2605.16941","url_pdf":"https://arxiv.org/pdf/2605.16941","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/2605.16941","atlas_url":null,"mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2605.16941"}},"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/Feng-Hong/WINO-DLLM","reach":null}],"summary":{"ran_draft_wrong":3,"ran_fixture":1,"ran":8,"unverified":1},"by_repo_kind":{"found_in_text":{"samples":13,"ran":12,"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":13,"samples":[{"code_sha256_prefix":"c3a6b977022957cb","entry":"Normalize","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"MMaDA/models/common_modules.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/MMaDA/models/common_modules.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"c3a6b977022957cb"}},{"code_sha256_prefix":"a5334721ab55f48f","entry":"add_gumbel_noise","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"LLaDA/decoding.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/LLaDA/decoding.py","link_basis":"plan_row","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"a5334721ab55f48f"}},{"code_sha256_prefix":"1f6ae43d06c76ba5","entry":"broadcast","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"MMaDA/models/misc.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/MMaDA/models/misc.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"1f6ae43d06c76ba5"}},{"code_sha256_prefix":"f9abf6d74699cc1c","entry":"causal_attention_bias","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"LLaDA/modeling_llada.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/LLaDA/modeling_llada.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"f9abf6d74699cc1c"}},{"code_sha256_prefix":"6aaeda12cec0718b","entry":"decoding_default","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"LLaDA/decoding.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/LLaDA/decoding.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"6aaeda12cec0718b"}},{"code_sha256_prefix":"a8366f238618edad","entry":"get_constant_schedule","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"MMaDA/models/lr_schedulers.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/MMaDA/models/lr_schedulers.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"a8366f238618edad"}},{"code_sha256_prefix":"11f07821d1b2286b","entry":"get_constant_schedule_with_warmup","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"MMaDA/models/lr_schedulers.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/MMaDA/models/lr_schedulers.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"11f07821d1b2286b"}},{"code_sha256_prefix":"1975e0c51e6f5753","entry":"get_linear_schedule_with_warmup","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"MMaDA/models/lr_schedulers.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/MMaDA/models/lr_schedulers.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"1975e0c51e6f5753"}},{"code_sha256_prefix":"6f22da9cc766ffbe","entry":"get_num_transfer_tokens","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"LLaDA/decoding.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/LLaDA/decoding.py","link_basis":"plan_row","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"6f22da9cc766ffbe"}},{"code_sha256_prefix":"3137073275f8c21a","entry":"nonlinearity","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"MMaDA/models/common_modules.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/MMaDA/models/common_modules.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"3137073275f8c21a"}},{"code_sha256_prefix":"0e7f77d6003e2412","entry":"parse_structured","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"MMaDA/models/misc.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/MMaDA/models/misc.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"0e7f77d6003e2412"}},{"code_sha256_prefix":"2d82abbef2c4addb","entry":"unpack_time","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"MMaDA/models/common_modules.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/MMaDA/models/common_modules.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"2d82abbef2c4addb"}},{"code_sha256_prefix":"376bf22d5737fc2a","entry":"get_logger","repo":"Feng-Hong/WINO-DLLM","repo_kind":"found_in_text","path":"MMaDA/models/logging.py","file_url":"https://github.com/Feng-Hong/WINO-DLLM/blob/HEAD/MMaDA/models/logging.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"376bf22d5737fc2a"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.CL","source":"arxiv_2026.jsonl"},"syntology_extracted_results":{"kind":"leaderboard_placements","source":"Syntology's leaderboard-shaped extractor over the paper's own arXiv-HTML tables: a model pointed at a cell, the number was read from that cell and checked against the board's metric, dataset, split and scale, and an independent check accepted the entry; not reviewed by the paper's authors or the archive's editors","extractor_model":"global.anthropic.claude-sonnet-4-5-20250929-v1:0","verifier_model":"us.anthropic.claude-sonnet-4-5-20250929-v1:0","prompt_sha":"fa63d4bb9d755694","coverage":{"sentence":"Syntology has checked 6,885 of the 9,623 papers on this site that are newer than the archive; results from the others appear after they are checked.","papers_newer_than_archive":9623,"papers_checked":6885},"entries":[],"not_placed":{"boards":4,"rejected_by_independent_check":3,"refused_by_a_rule":1,"check_did_not_answer":0,"proposed_without_a_cell":0,"declined_by_site":0}}}