{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,8]],"date-time":"2026-09-08T05:12:07Z","timestamp":1788844327798,"version":"build-2803163510"},"reference-count":103,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T00:00:00Z","timestamp":1782864000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T00:00:00Z","timestamp":1782864000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T00:00:00Z","timestamp":1782864000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"National Key Research and Development Program of China","award":["2024YFB4506000"],"award-info":[{"award-number":["2024YFB4506000"]}]},{"name":"National Natural Science Foundation of China","award":["61202091,62171155"],"award-info":[{"award-number":["61202091,62171155"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,7]]},"DOI":"10.1109\/icws72778.2026.00157","type":"proceedings-article","created":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T19:18:08Z","timestamp":1788808688000},"page":"1231-1242","source":"Crossref","is-referenced-by-count":0,"title":["Towards Efficient and Reliable On-Device Multi-Agent Systems: Challenges, Technologies and Explorations"],"prefix":"10.1109","author":[{"given":"Yanjun","family":"Shu","sequence":"first","affiliation":[{"name":"Harbin Institute of Technology","place":["Harbin, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Weiye","family":"Yan","sequence":"additional","affiliation":[{"name":"Harbin Institute of Technology","place":["Harbin, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhuangyu","family":"Ma","sequence":"additional","affiliation":[{"name":"Harbin Institute of Technology","place":["Harbin, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jian","family":"Yang","sequence":"additional","affiliation":[{"name":"Beijing Normal-Hong Kong Baptist University","place":["Zhuhai, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yanming","family":"Jiang","sequence":"additional","affiliation":[{"name":"China Mobile (Suzhou) Software Technology Co. Ltd.","place":["Suzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Min","family":"Li","sequence":"additional","affiliation":[{"name":"China Mobile (Suzhou) Software Technology Co. Ltd.","place":["Suzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tianyu","family":"Luo","sequence":"additional","affiliation":[{"name":"Harbin Institute of Technology","place":["Harbin, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Decheng","family":"Zuo","sequence":"additional","affiliation":[{"name":"Harbin Institute of Technology","place":["Harbin, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Multi-agent collaboration mechanisms: A survey of llms","author":"Tran","year":"2025","journal-title":"arXiv preprint"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2264"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1109\/ICWS67624.2025.00077"},{"key":"ref4","article-title":"Beyond self-talk: A communicationcentric survey of 1lm-based multi-agent systems","author":"Yan","year":"2025","journal-title":"arXiv preprint"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1109\/ICWS67624.2025.00136"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1109\/ICWS62655.2024.00070"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/ICWS67624.2025.00068"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/ICAIET65052.2025.11211133"},{"key":"ref9","article-title":"Mobile-agent: Autonomous multimodal mobile device agent with visual perception","author":"Wang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref10","article-title":"Learning latency-aware orchestration for parallel multi-agent systems","author":"Shi","year":"2026","journal-title":"arXiv preprint"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1145\/3706598.3713600"},{"key":"ref12","article-title":"Privatelora for efficient privacy preserving 1lm","author":"Wang","year":"2023","journal-title":"arXiv preprint"},{"key":"ref13","article-title":"Appcopilot: Toward general, accurate, long-horizon, and efficient mobile agent","author":"Fan","year":"2025","journal-title":"arXiv preprint"},{"key":"ref14","article-title":"Agent hospital: A simulacrum of hospital with evolvable medical agents","author":"Li","year":"2024","journal-title":"arXiv preprint"},{"key":"ref15","article-title":"A language agent for autonomous driving","author":"Mao","year":"2023","journal-title":"arXiv preprint"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/ICRA57147.2024.10611293"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1016\/j.comnet.2021.108176"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.3233\/web-190414"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1145\/3722565.3727192"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/TWC.2022.3191118"},{"key":"ref21","article-title":"Efficient on-device agents via adaptive context management","author":"Vijayvargiya","year":"2025","journal-title":"arXiv preprint"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1109\/ICAIBD69640.2026.11637248"},{"key":"ref23","article-title":"On the reliability limits of 1 lm based multi-agent planning","author":"Ao","year":"2026","journal-title":"arXiv preprint"},{"key":"ref24","article-title":"Smallthinker: A family of efficient large language models natively trained for local deployment","author":"Song","year":"2025","journal-title":"arXiv preprint"},{"key":"ref25","article-title":"Megrez2 technical report","author":"Li","year":"2025","journal-title":"arXiv preprint"},{"key":"ref26","article-title":"llama.cpp","volume-title":"ggml.ai","year":"2026"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1145\/3731569.3764808"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1145\/3669940.3707239"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1145\/3706418"},{"key":"ref30","article-title":"Llm as a system service on mobile devices","author":"Yin","year":"2024","journal-title":"arXiv preprint"},{"key":"ref31","volume-title":"Camphor: Collaborative agents for multi-input planning and high-order reasoning on device","author":"Fu","year":"2024"},{"key":"ref32","article-title":"Tokendance: Scaling multiagent llm serving via collective kv cache sharing","author":"Bian","year":"2026","journal-title":"arXiv preprint"},{"key":"ref33","article-title":"Agent memory below the prompt: Persistent q4 kv cache for multi-agent 1lm inference on edge devices","author":"Shkolnikov","year":"2026","journal-title":"arXiv preprint"},{"key":"ref34","article-title":"Oxygen: Unified kv cache management for vision-language-action models under multi-task parallelism","author":"Li","year":"2026","journal-title":"arXiv preprint"},{"key":"ref35","article-title":"Qkvshare: Quantized kv-cache handoff for multi-agent on-device llms","author":"Honavar","year":"2026","journal-title":"arXiv preprint"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.52202\/085713-0605"},{"key":"ref37","article-title":"Lragent: Efficient kv cache sharing for multi-lora 1lm agents","author":"Jeon","year":"2026","journal-title":"arXiv preprint"},{"key":"ref38","first-page":"10323","article-title":"SparseGPT: massive language models can be accurately pruned in one-shot","volume-title":"Proceedings of the 40th International Conference on Machine Learning, ser. ICML\u201923","volume":"202","author":"Frantar"},{"key":"ref39","first-page":"20336","article-title":"LoSparse: structured compression of large language models based on low-rank and sparse approximation","volume-title":"Proceedings of the 40th International Conference on Machine Learning","volume":"202","author":"Li","year":"2023"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v38i12.29237"},{"key":"ref41","first-page":"20852","article-title":"Less is more: task-aware layerwise distillation for language model compression","volume-title":"Proceedings of the 40th International Conference on Machine Learning, ser. ICML\u201923","volume":"202","author":"Liang"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.91"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.52202\/085713-3101"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.362"},{"key":"ref45","article-title":"A survey on inference optimization techniques for mixture of experts models","author":"Liu","year":"2025","journal-title":"arXiv preprint"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1145\/3575693.3575698"},{"key":"ref47","first-page":"31094","article-title":"FlexGen: high-throughput generative inference of large language models with a single GPU","volume-title":"Proceedings of the 40th International Conference on Machine Learning","author":"Sheng","year":"2023"},{"key":"ref48","article-title":"Fast inference of mixture-of-experts language models with offloading","author":"Eliseev","year":"2023","journal-title":"arXiv preprint"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.3390\/electronics13112077"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1145\/3676536.3676741"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1109\/TMC.2025.3546466"},{"key":"ref52","article-title":"MoE-Infinity: Efficient MoE inference on personal machines with sparsity-aware expert cache","author":"Xue","year":"2024","journal-title":"arXiv preprint"},{"key":"ref53","article-title":"fMoE: Fine-grained expert offloading for large mixture-of-experts serving","author":"Yu","year":"2025","journal-title":"arXiv preprint"},{"key":"ref54","article-title":"HOBBIT: A mixed precision expert offloading system for fast MoE inference","author":"Tang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.1109\/DAC63849.2025.11133274"},{"key":"ref56","article-title":"SliceMoE: Bit-sliced expert caching under miss-rate constraints for efficient MoE inference","author":"Choi","year":"2025","journal-title":"arXiv preprint"},{"key":"ref57","article-title":"BuddyMoE: Exploiting expert redundancy to accelerate memory-constrained mixture-of-experts inference","author":"Wang","year":"2025","journal-title":"arXiv preprint"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1145\/3680207.3723493"},{"key":"ref59","article-title":"ProMoE: Fast MoE-based LLM serving using proactive caching","author":"Song","year":"2024","journal-title":"arXiv preprint"},{"key":"ref60","article-title":"FloE: On-the-fly MoE inference on memory-constrained GPU","author":"Zhou","year":"2025","journal-title":"arXiv preprint"},{"key":"ref61","article-title":"PreScope: Unleashing the power of prefetching for resource-constrained MoE inference","author":"Yu","year":"2025","journal-title":"arXiv preprint"},{"key":"ref62","article-title":"ExpertFlow: Optimized expert activation and token allocation for efficient mixture-of-experts inference","author":"He","year":"2024","journal-title":"arXiv preprint"},{"key":"ref63","article-title":"Llm-based agents suffer from hallucinations: A survey of taxonomy, methods, and directions","author":"Lin","year":"2025","journal-title":"arXiv preprint"},{"key":"ref64","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.ijcnlp-main.20"},{"key":"ref65","article-title":"Saup: Situation awareness uncertainty propagation on 1lm agent","author":"Zhao","year":"2024","journal-title":"arXiv preprint"},{"key":"ref66","article-title":"Improving factuality and reasoning in language models through multiagent debate","volume-title":"Forty-first international conference on machine learning","author":"Du","year":"2024"},{"key":"ref67","article-title":"Crab: A semantics-aware checkpoint\/restore runtime for agent sandboxes","author":"Wu","year":"2026","journal-title":"arXiv preprint"},{"key":"ref68","article-title":"Stop wasting your tokens: Towards efficient runtime multi-agent systems","author":"Lin","year":"2025","journal-title":"arXiv preprint"},{"issue":"3","key":"ref69","first-page":"1","article-title":"Edge ai: a review of machine learning models for resource-constrained devices","volume":"5","author":"Kumar","year":"2024","journal-title":"Artificial Intelligence and Machine Learning Review"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.52202\/085713-4082"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1613"},{"key":"ref72","article-title":"Self-consistency improves chain of thought reasoning in language models","author":"Wang","year":"2022","journal-title":"arXiv preprint"},{"key":"ref73","first-page":"9079","article-title":"Chateval: Towards better llm-based evaluators through multi-agent debate","volume-title":"International conference on learning representations","volume":"2024","author":"Chan","year":"2024"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.992"},{"key":"ref75","doi-asserted-by":"publisher","DOI":"10.52202\/085713-3405"},{"key":"ref76","article-title":"Orchestrating intelligence: Confidenceaware routing for efficient multi-agent collaboration across multi-scale models","author":"Wang","year":"2026","journal-title":"arXiv preprint"},{"key":"ref77","article-title":"Should we be going mad? a look at multi-agent debate strategies for llms","author":"Smit","year":"2023","journal-title":"arXiv preprint"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.65109\/dsdx8860"},{"key":"ref79","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-emnlp.427"},{"key":"ref80","article-title":"Chateval: Towards better llm-based evaluators through multi-agent debate","author":"Chan","year":"2023","journal-title":"arXiv preprint"},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2026.findings-acl.1694"},{"key":"ref82","article-title":"Epistemic gain, aleatoric cost: Uncertainty decomposition in multi-agent debate for math reasoning","author":"Qiao","year":"2026","journal-title":"arXiv preprint"},{"key":"ref83","doi-asserted-by":"publisher","DOI":"10.65109\/CBJO8409"},{"key":"ref84","article-title":"Can llm agents really debate? a controlled study of multi-agent debate in logical reasoning","author":"Wu","year":"2025","journal-title":"arXiv preprint"},{"key":"ref85","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-acl.495"},{"key":"ref86","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2279"},{"key":"ref87","article-title":"Dynadebate: Breaking homogeneity in multi-agent debate with dynamic path generation","author":"Li","year":"2026","journal-title":"arXiv preprint"},{"key":"ref88","article-title":"Where llm agents fail and how they can learn from failures","author":"Zhu","year":"2025","journal-title":"arXiv preprint"},{"key":"ref89","article-title":"React: Synergizing reasoning and acting in language models","author":"Yao","year":"2022","journal-title":"arXiv preprint"},{"key":"ref90","article-title":"Acrfence: Preventing semantic rollback attacks in agent checkpoint-restore","author":"Zheng","year":"2026","journal-title":"arXiv preprint"},{"key":"ref91","doi-asserted-by":"publisher","DOI":"10.1109\/TCCN.2025.3528892"},{"key":"ref92","article-title":"Mas-fire: Fault injection and reliability evaluation for 1lm-based multi-agent systems","author":"Jia","year":"2026","journal-title":"arXiv preprint"},{"key":"ref93","doi-asserted-by":"publisher","DOI":"10.1145\/3749421.3749436"},{"key":"ref94","article-title":"Get experience from practice: Llm agents with record & replay","author":"Feng","year":"2025","journal-title":"arXiv preprint"},{"key":"ref95","doi-asserted-by":"publisher","DOI":"10.1145\/3805621.3807638"},{"key":"ref96","article-title":"Fault-tolerant sandboxing for ai coding agents: A transactional approach to safe autonomous execution","author":"Yan","year":"2025","journal-title":"arXiv preprint"},{"key":"ref97","article-title":"Shielda: Structured handling of exceptions in 1lm-driven agentic workflows","author":"Zhou","year":"2025","journal-title":"arXiv preprint"},{"key":"ref98","article-title":"Autogen: Enabling next-gen llm applications via multi-agent conversations","volume-title":"First conference on language modeling","author":"Wu","year":"2024"},{"key":"ref99","first-page":"80569","article-title":"Agent-as-a-judge: Evaluate agents with agents","volume-title":"International Conference on Machine Learning. PMLR","author":"Zhuge","year":"2025"},{"key":"ref100","article-title":"Auto-eval judge: Towards a general agentic framework for task completion evaluation","author":"Bhonsle","year":"2025","journal-title":"arXiv preprint"},{"key":"ref101","doi-asserted-by":"publisher","DOI":"10.1147\/sj.52.0078"},{"key":"ref102","volume-title":"\u2019smolagents\u2019: a smol library to build great agentic systems","author":"Roucher","year":"2025"},{"key":"ref103","article-title":"Gaia: a benchmark for general ai assistants","volume-title":"The Twelfth International Conference on Learning Representations","author":"Mialon","year":"2023"}],"event":{"name":"2026 IEEE International Conference on Web Services (ICWS)","location":"Sydney, Australia","start":{"date-parts":[[2026,7,13]]},"end":{"date-parts":[[2026,7,18]]}},"container-title":["2026 IEEE International Conference on Web Services (ICWS)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11670018\/11670022\/11670154.pdf?arnumber=11670154","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,8]],"date-time":"2026-09-08T04:53:29Z","timestamp":1788843209000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11670154\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7]]},"references-count":103,"URL":"https:\/\/doi.org\/10.1109\/icws72778.2026.00157","relation":{},"subject":[],"published":{"date-parts":[[2026,7]]}}}