[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-238628-105":59,"doc-detail-238628-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","karma-adaptive-video-streaming-via-causal-sequence-modeling","Karma - Adaptive Video Streaming via Causal Sequence Modeling","","Optimal adaptive bitrate (ABR) decisions require modeling state transitions that jointly reflect time-varying environmental observations, historical returns, and chosen actions. Existing learning-based ABR methods often rely only on past observations for the next action, which can trigger chained deviations under unfamiliar network conditions and reduce generalization. Karma improves generalization through causal sequence modeling, maintaining multi-dimensional history of observations, returns, and actions and using a decision-transformer to refine actions when deviations occur. Evaluations with trace-driven simulations and real-world field tests show 10.8%–18.7% average QoE gains and strong unseen-network performance.",{"@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/karma-adaptive-video-streaming-via-causal-sequence-modeling/238628/",{"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/karma-adaptive-video-streaming-via-causal-sequence-modeling/238628.png","ImageObject",300,407,{"name":92,"@type":93},"Ava Thompson","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-01","2026-09-11",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 Karma address in adaptive bitrate (ABR) learning methods?","Question",{"text":112,"@type":113},"Karma targets the generalization loss caused by observation-only policies, which can lead to chained deviations from optimal actions when encountering unfamiliar network observations.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"How does Karma differ from direct observation-to-action ABR mapping?",{"text":117,"@type":113},"Karma keeps a multi-dimensional time series of observations, returns, and actions and uses causal sequence modeling via a decision transformer to compute the next action.",{"name":119,"@type":110,"acceptedAnswer":120},"How is the return signal constructed in Karma?",{"text":121,"@type":113},"Karma uses maximum cumulative future QoE (QoE-to-go) as an extended return signal, periodically estimated from current network conditions and playback status.","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},238628,1789138545,{"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":44,"language":139,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":140,"faqs":141,"seo_title":142,"seo_description":67,"update_tm":129,"read_time":143,"preview_page_count":4},1649267921044,"https://us-avatar.wpscdn.com/avatar/1800007509477c92dfb?_k=1786009248482753345","Karma: Adaptive Video Streaming via Causal Sequence Modeling  \nBowei Xu  \nNanjing University Nanjing, China [xubowei@smail.nju.edu.cn](xubowei@smail.nju.edu.cn)  \nHao Chen  \nNanjing University Nanjing, China [chenhao1210@nju.edu.cn](chenhao1210@nju.edu.cn)  \nZhan Ma  \nNanjing University Nanjing, China [mazhan@nju.edu.cn](mazhan@nju.edu.cn)  \nABSTRACT  \nOptimal adaptive bitrate (ABR) decision depends on a comprehensive characterization of state transitions that involve interrelated modalities over time including environmental observations, returns, and actions. However, state-of-the-art learning-based ABR algorithms solely rely on past observations to decide the next action. This paradigm tends to cause a chain of deviations from optimal action when encountering unfamiliar observations, which consequently undermines the model generalization.  \nThis paper presents Karma, an ABR algorithm that utilizes causal sequence modeling to improve generalization by comprehending the interrelated causality among past observations, returns, and actions and timely refining action when deviation occurs. Unlike direct observation-to-action mapping, Karma recurrently maintains a multi-dimensional time series of observations, returns, and actions as input and employs causal sequence modeling via a decision transformer to determine the next action. In the input sequence, Karma uses the maximum cumulative future quality of experience (QoE)(a.k.a, QoE-to-go) as an extended return signal, which is periodically estimated based on current network conditions and playback status. We evaluate Karma through trace-driven simulations and real-world field tests, demonstrating superior performance compared to existing state-of-the-art ABR algorithms, with an average QoE improvement ranging from 10.8% to 18.7% across diverse network conditions. Furthermore, Karma exhibits strong generalization capabilities, showing leading performance under unseen networks in both simulations and real-world tests.  \nCCS CONCEPTS  \n• Information systems → Multimedia streaming; Information systems applications.  \nKEYWORDS  \nSequence Modeling, Decision Transformer, Adaptive BitRate, Video Streaming  \nACM Reference Format:  \nBowei Xu, Hao Chen, and Zhan Ma. 2023. Karma: Adaptive Video Streaming via Causal Sequence Modeling. In Proceedings of the 31stACM International Conference on Multimedia (MM’23) . ACM, New York, NY, USA, 9 pages. [https://doi.org/10.1145/3581783.3612177](https://doi.org/10.1145/3581783.3612177)  \nPermission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission [and/or a fee. Request permissions from permissions@acm.org](and/or a fee. Request permissions from permissions@acm.org).  \nMM’23, October 29-November 3, 2023, Ottawa, ON, Canada  \n© 2023 Copyright held by the owner/author(s) . Publication rights licensed to ACM. ACM ISBN 979-8-4007-0108-5/23/10. . . $15.00  \n[https://doi.org/10.1145/3581783.3612177](https://doi.org/10.1145/3581783.3612177)  \n(a) Existing schemes via observation-to-action mapping  \n(b) Karma via causal sequence modeling  \nFigure 1: ABR Decision In Use : Karma vs. existing algorithms.  \n1 INTRODUCTION  \nIn recent years, a remarkable surge of [HTTP-based](HTTP-based) video traffic [9, 21] has been propelled by the proliferation of video streaming applications. Adaptive Bitrate (ABR) algorithms [6, 12, 28] have emerged as prominent tools and have been employed by content providers to optimize the video quality of streaming services. Typically implemented on the client-side video player, th","cbCaipCWyUb5nynW","https://ap.wps.com/l/cbCaipCWyUb5nynW","pdf",1416352,"English","# Introduction\n## Adaptive bitrate and generalization challenge\n# Proposed method: Karma\n## Causal sequence modeling with decision transformer\n## QoE-to-go as extended return signal\n# Evaluation\n## Trace-driven simulations\n## Real-world field tests","[{\"question\":\"What problem does Karma address in adaptive bitrate (ABR) learning methods?\",\"answer\":\"Karma targets the generalization loss caused by observation-only policies, which can lead to chained deviations from optimal actions when encountering unfamiliar network observations.\"},{\"question\":\"How does Karma differ from direct observation-to-action ABR mapping?\",\"answer\":\"Karma keeps a multi-dimensional time series of observations, returns, and actions and uses causal sequence modeling via a decision transformer to compute the next action.\"},{\"question\":\"How is the return signal constructed in Karma?\",\"answer\":\"Karma uses maximum cumulative future QoE (QoE-to-go) as an extended return signal, periodically estimated from current network conditions and playback status.\"}]","Karma - Adaptive Video Streaming via Causal Sequence Modeling | PDF",23]