[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-83965-en":3,"doc-seo-83965-105":30,"detail-sidebar-cat-0-en-105":92},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},83965,687197207639,"Asher","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","Decision Protocols in Multi-Agent Large Language Model Conversations","Improving Large Language Model (LLM) task performance faces diminishing returns and high costs when scaling. Multi-Agent Systems (MAS) mitigate training cost by distributing work among specialized agents, trading it for increased test-time due to discussion and decision-making. The decision protocol is central to MAS, defining how agents collaborate to form a final solution. The thesis presents the Multi-Agent LLM (MALLM) framework, implementing and evaluating voting, consensus, and judge decision mechanisms across knowledge and logic tasks, comparing performance, answer diversity, and protocol variations.","arXiv :2607 .05477v 1 [ cs .MA] 6 Jul 2026  \nMaster’s Thesis  \nDecision Protocols in Multi-Agent Large Language  \nModel Conversations  \nLars Benedikt Kaesberg  \nIn Fulfillment of the Requirements  \nfor the Degree of  \nMaster of Science  \nMain Examiner: Prof. Dr. Bela Gipp  \nSecond Examiner: Dr. Terry Lima Ruas  \nSupervisor: Jan Philip Wahle  \nGippLab  \nGeorg August University of G¨ottingen  \nG¨ottingen, December 2024  \nContents  \n1. Introduction 1  \n2. Related Work 5  \n2.1. Decision-Making . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5  \n2.2. LLMs and Agents ..................................... 6  \n2.3. Multi-Agent Frameworks ................................. 7  \n2.4. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8  \n3. Methodology 9  \n3.1. Multi-Agent Discussion Framework ........................... 9  \n3.2. Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15  \n4. Experiments 17  \n4.1. Performance of Decision Protocols . . . . . . . . . . . . . . . . . . . . . . . . . . . 19  \n4.2. Answer Diversity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24  \n4.3. Alterations in Decision Protocols . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28  \n4.4. Challenge of Final Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31  \n4.5. Compute Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33  \n5. Epilogue 35  \n5.1. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35  \n5.2. Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36  \n5.3. Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37  \n6. References 38  \nA. Results i  \nB. Prompts xi  \nC. MALLM Setup xv  \nD. Example Discussions xxiv  \nE. AI Usage Card xli  \nI hereby declare that I have written this thesis independently without any help from others and without the use of documents or aids other than those stated. I have mentioned all used sources and cited them correctly according to established academic citation rules.  \nHelsinki, December 31th, 2024  \nAbstract  \nImproving the task performance of Large Language Models (LLMs) is essential, yet scaling these models faces significant challenges such as diminishing returns and high costs. Multi-Agent Systems (MAS) offer a promising solution by distributing tasks among specialized agents to improve the overall task performance. This can reduce training costs at the expense of increased test time due to the discussion and decisionmaking process. The decision protocol is a critical component of MAS because it specifies how multiple agents collaborate to create a final solution. This thesis introduces the Multi-Agent LLM (MALLM) framework, which implements and evaluates various decision protocols, namely voting, consensus, and judge decision mechanisms, to simulate multi-agent discussions for conversational task solving. Unlike previous work that used a single decision protocol or tested them on limited datasets, this study systematically examines their impact on a diverse set of tasks, ranging from knowledge-based datasets (MMLU, MMLU-Pro, GPQA) and logic-based datasets (StrategyQA, MuSR, Math-lvl-5, SQuAD 2.0) . The results indicate that consensus protocols excel in knowledge-intensive domains while voting and judge protocols are more effective for logic-based tasks.  \nIncreasing response diversity through independent solution generation improves decision quality, while changes in information access during the decision process have minimal impact.  \nKeywords: Large Language Models, Multi-Agent Systems, Decision Protocols, TestTime Compute, Reasoning  \n1. Introduction  \nRecent years have shown a trend to scale the size of LLMs to improve task performance [28] . It is unclear how far this trend can be pushed, as a furthe","cbCaipDRhnCrd8aF","https://ap.wps.com/l/cbCaipDRhnCrd8aF","pdf",1089990,7,1,86,"English","en",105,"# Introduction\n# Related Work\n# Methodology\n# Experiments\n## Performance of Decision Protocols\n## Answer Diversity\n## Alterations in Decision Protocols\n## Challenge of Final Solution\n## Compute Time\n# Epilogue\n## Conclusion\n## Future Work\n## Limitations\n# References","[{\"question\":\"What problem does the thesis address regarding scaling LLMs?\",\"answer\":\"Scaling LLMs for higher task performance can show diminishing returns and incurs high costs. The thesis motivates alternatives that improve performance without continually larger models.\"},{\"question\":\"What is the Multi-Agent LLM (MALLM) framework?\",\"answer\":\"MALLM implements and evaluates multi-agent decision protocols for conversational task solving. It simulates multi-agent discussions using voting, consensus, and judge mechanisms.\"},{\"question\":\"How do decision protocols differ in effectiveness across task types?\",\"answer\":\"Results show consensus protocols work best in knowledge-intensive domains, while voting and judge protocols perform better on logic-based tasks. This indicates protocol choice should depend on the task category.\"}]",1784191704,217,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"decision-protocols-in-multi-agent-large-language-model-conversations","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/decision-protocols-in-multi-agent-large-language-model-conversations/83965/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-07-27","2026-07-16",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"What problem does the thesis address regarding scaling LLMs?","Question",{"text":76,"@type":77},"Scaling LLMs for higher task performance can show diminishing returns and incurs high costs. The thesis motivates alternatives that improve performance without continually larger models.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What is the Multi-Agent LLM (MALLM) framework?",{"text":81,"@type":77},"MALLM implements and evaluates multi-agent decision protocols for conversational task solving. It simulates multi-agent discussions using voting, consensus, and judge mechanisms.",{"name":83,"@type":74,"acceptedAnswer":84},"How do decision protocols differ in effectiveness across task types?",{"text":85,"@type":77},"Results show consensus protocols work best in knowledge-intensive domains, while voting and judge protocols perform better on logic-based tasks. This indicates protocol choice should depend on the task category.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":46,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":107,"slug":138},19,"General","general"]