[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-504714-105":59,"doc-detail-504714-en":130},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","alphaone-reasoning-models-thinking-slow-and-fast-at-test-time","ALPHAONE: Reasoning Models Thinking Slow and Fast at Test Time","","ALPHAONE (α1) proposes a universal test-time framework to modulate reasoning progress in large reasoning models. The method introduces an α moment that marks a scaled thinking phase, then schedules slow-to-fast transitions by modeling reasoning transition token insertion as a Bernoulli stochastic process. After the α moment, α1 deterministically ends slow thinking via an end-of-thinking token. Experiments on mathematical, coding, and scientific benchmarks show improved reasoning capability and efficiency.",{"@graph":69,"@context":122},[70,84,105],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/alphaone-reasoning-models-thinking-slow-and-fast-at-test-time/504714/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/alphaone-reasoning-models-thinking-slow-and-fast-at-test-time/504714.png","ImageObject",300,407,{"name":92,"@type":93},"Levi","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-17","2026-10-04",true,{"@type":102,"interactionType":103,"userInteractionCount":34},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What problem does ALPHAONE (α1) address at test time?","Question",{"text":112,"@type":113},"It addresses how to reliably modulate reasoning progress in large reasoning models during inference, improving the transition between slow and fast thinking for complex problems.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"How does α1 implement slow-to-fast reasoning scheduling?",{"text":117,"@type":113},"α1 models the insertion of reasoning transition tokens as a Bernoulli stochastic process within a scaled thinking phase before the α moment, enabling dynamic slow-thinking transitions.",{"name":119,"@type":110,"acceptedAnswer":120},"What happens after the α moment in α1?",{"text":121,"@type":113},"After the α moment, α1 deterministically terminates slow thinking using an end-of-thinking token, which supports faster reasoning and efficient answer generation.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},504714,1791453017,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":34,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":139,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":144,"read_time":145},7971461740909,"https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d","ALPHAONE: Reasoning Models Thinking Slow and Fast at Test Time  \nJunyu Zhang† Runpei Dong† Han Wang Xuying Ning  \nHaoran Geng Peihao Li Xialin He Yutong Bai Jitendra Malik Saurabh Gupta Huan Zhang  \nUniversity of Illinois Urbana-Champaign UC Berkeley †Equal contributions. Correspondence: {junyuz6, runpeid2, [huanz}@illinois.edu](huanz}@illinois.edu)  \narXiv :2505 .24863v1 [ cs .CL] 30 May 2025  \nAbstract  \nThis paper presents ALPHAONE (α1), a universal framework for modulating reasoning progress in large reasoning models (LRMs) at test time. α1 first introduces α moment, which represents the scaled thinking phase with a universal parameter α . Within this scaled pre-α moment phase, it dynamically schedules slow thinking transitions by modeling the insertion of reasoning transition tokens as a Bernoulli stochastic process. After the α moment, α1 deterministically terminates slow thinking with the end-of-thinking token, thereby fostering fast reasoning and efficient answer generation. This approach unifies and generalizes existing monotonic scaling methods by enabling flexible and dense slow-to-fast reasoning modulation. Extensive empirical studies on various challenging benchmarks across mathematical, coding, and scientific domains demonstrate α1’s superior reasoning capability and efficiency. Project page:  \n[https://alphaone-project.github.io/](https://alphaone-project.github.io/)  \n1 Introduction  \n“The most effortful forms of slow thinking are those that require you to think fast.”  \nThinking, Fast and Slow (Kahneman, 2011)  \nLarge Reasoning Models (LRMs) such as OpenAI o1 (Jaech et al., 2024) and DeepSeekR1 (DeepSeek-AI et al., 2025) have demonstrated unprecedented progress in approaching humanlike system-2 reasoning capabilities, enabling slow thinking—slowing down reasoning progress 1 attest time—for solving complex reasoning problems that require high-order cognitive processing. These advanced models are trained to utilize slow thinking via reinforcement learning, enabling LRMs  \n1Consider reasoning progress as a metric ranging from 0 to 1, indicating the start and the end of reasoning, respectively. It increases slowly or fast at the pace of thinking. See Fig. 1 and Section 2 for a more illustrative and detailed explanation.  \nStart 􀀡 Moment Reasoning Time  \nFigure 1: Conceptual illustration of reasoning modulation strategies. Our α1 employs a slow-to-fast reasoning schedule controlled by α . α1 scales more efficiently than monotonously increasing method s1 (yellow) and generally outperforms monotonously decreasing (purple) approaches.  \nto slow down reasoning progress automatically. Is such automatic slowing down of reasoning progress determined by LRMs sufficiently reliable? According to Kahneman (2011), humans typically think fast first and activate slow thinking when running into difficulty, through a conscious control of system-1-to-2 reasoning transitioning, resulting in overall comprehensive but efficient reasoning. While similar to human systems and interesting results have been observed, a lot of works have pointed out that the LRMs themselves are prone to overthinking (Chen et al., 2024b ; Sui et al., 2025 ; Pu et al., 2025 ; Yang et al., 2025c) or underthinking (Su et al., 2025 ; Yang et al., 2025d ; Wang et al., 2025) . This is because of the inability of LRMs to find the optimal human-like system-1-to-2 reasoning transitioning and limited reasoning capabilities, leading to unsatisfactory reasoning performance.  \nTo overcome this limitation, existing works scale LRMs at test time in mainly two ways. i) paral-  \nQuestion  \nLet 􀀡 be the number of sequences 􀀢! , 􀀢\" ,…, 􀀢\\# .…What is the remainder when 􀀡 is divided by 10?  \nVanilla  \nOkay, so I need to figure out … \\n\\n First, let me parse the problem.… Then,…\\n\\n Wait, but that can't be, because… \u003C/think> … So, the  \nremainder when K is divided by 10 is 0.  Wait Frequency: 23  \n1  \n0  \nFirst, let me parse the problem…\\n\\n  \nWait, actually, reading the problem ","cbCaifM9YesxYfru","https://ap.wps.com/l/cbCaifM9YesxYfru","pdf",1974880,23,"English","# Abstract\n# Introduction\n## Thinking, Fast and Slow\n## Large Reasoning Models and Limitations\n## Test-Time Scaling Approaches","[{\"question\":\"What problem does ALPHAONE (α1) address at test time?\",\"answer\":\"It addresses how to reliably modulate reasoning progress in large reasoning models during inference, improving the transition between slow and fast thinking for complex problems.\"},{\"question\":\"How does α1 implement slow-to-fast reasoning scheduling?\",\"answer\":\"α1 models the insertion of reasoning transition tokens as a Bernoulli stochastic process within a scaled thinking phase before the α moment, enabling dynamic slow-thinking transitions.\"},{\"question\":\"What happens after the α moment in α1?\",\"answer\":\"After the α moment, α1 deterministically terminates slow thinking using an end-of-thinking token, which supports faster reasoning and efficient answer generation.\"}]","ALPHAONE: Reasoning Models Thinking Slow and Fast at Test Time | PDF",1791073830,58]