{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T08:01:28.298Z","headline":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","description":"哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","url":"https://www.aioga.com/news/cmrvhhxue028ibihbb26yqn5a/","mainEntityOfPage":"https://www.aioga.com/news/cmrvhhxue028ibihbb26yqn5a/","datePublished":"2026-07-21T00:00:00.000Z","dateModified":"2026-07-21T00:00:00.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://arxiv.org/abs/2607.18955","https://aihot.virxact.com/items/cmrvhhxue028ibihbb26yqn5a"],"canonicalUrl":"https://www.aioga.com/news/cmrvhhxue028ibihbb26yqn5a/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。 Aioga 将其归入「论文研究」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/cmrvhhxue028ibihbb26yqn5a/","dateCreated":"2026-07-21T00:00:00.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"arXiv source article","url":"https://arxiv.org/abs/2607.18955","datePublished":"2026-07-21T00:00:00.000Z","provider":{"@type":"Organization","name":"arXiv","url":"https://arxiv.org/abs/2607.18955"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrvhhxue028ibihbb26yqn5a","datePublished":"2026-07-21T00:00:00.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrvhhxue028ibihbb26yqn5a"}}],"aggregationSource":"HuggingFace Daily Papers（社区热门论文）","originalPublisher":{"name":"arXiv","url":"https://arxiv.org/abs/2607.18955"},"article":{"id":"cmrvhhxue028ibihbb26yqn5a","slug":"cmrvhhxue028ibihbb26yqn5a","url":"https://www.aioga.com/news/cmrvhhxue028ibihbb26yqn5a/","title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","title_en":"H^2SD： Hybrid Hindsight Self-Distillation","summary":"哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","source":"HuggingFace Daily Papers（社区热门论文）","sourceUrl":"https://arxiv.org/abs/2607.18955","aiHotUrl":"https://aihot.virxact.com/items/cmrvhhxue028ibihbb26yqn5a","publishedAt":"2026-07-21T00:00:00.000Z","category":"论文研究","score":50,"selected":false,"articleBody":["arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.","Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs ：https://info.arxiv.org/labs/index.html."],"articleImages":[{"sourceUrl":"https://arxiv.org/static/base/1.0.1/images/funders/simons-foundation.png","alt":"Simons Foundation","afterParagraph":1,"url":"/media/articles/cmrvhhxue028ibihbb26yqn5a/e2d7f38d62f5ca91.png"},{"sourceUrl":"https://arxiv.org/static/base/1.0.1/images/funders/simons-foundation-international.png","alt":"Simons Foundation International","afterParagraph":1,"url":"/media/articles/cmrvhhxue028ibihbb26yqn5a/1d56e29c5557cbdc.png"},{"sourceUrl":"https://arxiv.org/static/base/1.0.1/images/funders/schmidt-sciences.png","alt":"Schmidt Sciences","afterParagraph":1,"url":"/media/articles/cmrvhhxue028ibihbb26yqn5a/8e18212b8219c104.png"}],"mediaStatus":"ok","articleBodyZh":["arXivLabs 是一个框架，允许合作者直接在我们的网站上开发和分享新的 arXiv 功能。","有一个可以为 arXiv 社区增加价值的项目想法吗？了解更多关于 arXivLabs 的信息：https://info.arxiv.org/labs/index.html。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。 Aioga 将其归入「论文研究」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：模型与研究类动态需要结合能力边界、开放方式、成本、可用性和真实任务表现判断，单项指标领先不等于已经形成稳定采用。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察官方文档、实际可用性、价格变化、开发者反馈和竞品回应。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-07-23T08:10:16.743Z","sourceHash":"4a55242636ec980a","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["论文研究","HuggingFace Daily Papers（社区热门论文）"],"translations":{"zh-CN":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga AI资讯","description":"哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","url":"https://www.aioga.com/news/cmrvhhxue028ibihbb26yqn5a/"},"en":{"title":"H^2SD： Hybrid Hindsight Self-Distillation","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under Research. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"Research","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H^2SD： Hybrid Hindsight Self-Distillation - Aioga AI News","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under Research. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RL...","url":"https://www.aioga.com/en/news/cmrvhhxue028ibihbb26yqn5a/"},"ja":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aiogaは「論文研究」の動きとして、HuggingFace Daily Papers（社区热门论文） からの更新を追跡しています。哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"論文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga AIニュース","description":"Aiogaは「論文研究」の動きとして、HuggingFace Daily Papers（社区热门论文） からの更新を追跡しています。哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基...","url":"https://www.aioga.com/ja/news/cmrvhhxue028ibihbb26yqn5a/"},"ko":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga는 HuggingFace Daily Papers（社区热门论文）의 업데이트를 연구 흐름으로 추적합니다. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"연구","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga AI 뉴스","description":"Aioga는 HuggingFace Daily Papers（社区热门论文）의 업데이트를 연구 흐름으로 추적합니다. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时...","url":"https://www.aioga.com/ko/news/cmrvhhxue028ibihbb26yqn5a/"},"es":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga sigue esta actualización de HuggingFace Daily Papers（社区热门论文） dentro de Investigación. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"Investigación","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga Noticias de IA","description":"Aioga sigue esta actualización de HuggingFace Daily Papers（社区热门论文） dentro de Investigación. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基...","url":"https://www.aioga.com/es/news/cmrvhhxue028ibihbb26yqn5a/"},"fr":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga suit cette mise à jour de HuggingFace Daily Papers（社区热门论文） dans la catégorie Recherche. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"Recherche","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga Actualités IA","description":"Aioga suit cette mise à jour de HuggingFace Daily Papers（社区热门论文） dans la catégorie Recherche. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推...","url":"https://www.aioga.com/fr/news/cmrvhhxue028ibihbb26yqn5a/"},"de":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga KI-News","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/de/news/cmrvhhxue028ibihbb26yqn5a/"},"pt-BR":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga Notícias de IA","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/pt-BR/news/cmrvhhxue028ibihbb26yqn5a/"},"ru":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga Новости ИИ","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/ru/news/cmrvhhxue028ibihbb26yqn5a/"},"ar":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga أخبار الذكاء الاصطناعي","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/ar/news/cmrvhhxue028ibihbb26yqn5a/"},"hi":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga AI समाचार","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/hi/news/cmrvhhxue028ibihbb26yqn5a/"},"it":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga Notizie IA","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/it/news/cmrvhhxue028ibihbb26yqn5a/"},"nl":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga AI-nieuws","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/nl/news/cmrvhhxue028ibihbb26yqn5a/"},"tr":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga AI Haberleri","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/tr/news/cmrvhhxue028ibihbb26yqn5a/"},"vi":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Tin tức AI Aioga","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/vi/news/cmrvhhxue028ibihbb26yqn5a/"},"id":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Berita AI Aioga","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/id/news/cmrvhhxue028ibihbb26yqn5a/"},"th":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - ข่าว AI Aioga","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/th/news/cmrvhhxue028ibihbb26yqn5a/"},"pl":{"title":"H2SD：混合事后自我蒸馏框架提升大模型推理能力","summary":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、OPSD和RLSD基线，同时保持稳定优化与生成效率。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"H2SD：混合事后自我蒸馏框架提升大模型推理能力 - Aioga Wiadomości AI","description":"Aioga tracks this update from HuggingFace Daily Papers（社区热门论文） under 论文研究. 哈尔滨工业大学团队提出混合事后自我蒸馏框架H2SD，根据轨迹正确性差异化使用教师信号：成功轨迹仅用教师概率调节更新幅度，失败轨迹则通过反向KL散度提供显式分布修正。在多个推理基准上，H2SD持续优于RLVR、O...","url":"https://www.aioga.com/pl/news/cmrvhhxue028ibihbb26yqn5a/"}}}}