Hugging Face Daily Papers · · 5 min read

What Should Agents Say? Action-state Communication for Efficient Multi-Agent Systems

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

Multi-agent systems (MAS) built on large language models are typically organized around roles, pipelines, and turn schedules, while the content that agents pass to one another is often left as unconstrained natural language. However, this free-form communication can rapidly inflate token usage, consume the shared context window, and ultimately affect both system performance and inference cost. We analyze five common inter-agent communication strategies across two MAS topologies, finding that no fixed strategy is universally optimal. Instead, effective inter-agent messages consistently preserve action-centered information needed by downstream agents. Building on this, we propose the PACT (Protocolized Action-state Communication and Transmission), which treats inter-agent communication as a public state-update problem and projects each raw agent output into a compact action-state record before it enters shared history. Across different MAS topologies, PACT consistently improves the performance-cost trade-off, achieving comparable or stronger task performance with substantially fewer tokens. The gains extend to production coding harnesses: PACT lifts OpenHands' resolve rate at -10% tokens-per-resolved, and is resolve-neutral on SWE-agent while halving input tokens. Our code is publicly available at <a href=\"https://github.com/iNLP-Lab/PACT\" rel=\"nofollow\">https://github.com/iNLP-Lab/PACT</a> .</p>\n","updatedAt":"2026-06-10T01:39:34.784Z","author":{"_id":"65d7b983baa72790a1151923","avatarUrl":"/avatars/938531e84ca01a0c5a2a174057e3e9c5.svg","fullname":"Chen Huang","name":"Albus-Chen","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":2,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9031363725662231},"editors":["Albus-Chen"],"editorAvatarUrls":["/avatars/938531e84ca01a0c5a2a174057e3e9c5.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.05304","authors":[{"_id":"6a28c001e7d78ea7587e5296","name":"Chen Huang","hidden":false},{"_id":"6a28c001e7d78ea7587e5297","name":"Yuhao Wu","hidden":false},{"_id":"6a28c001e7d78ea7587e5298","name":"Wenxuan Zhang","hidden":false}],"publishedAt":"2026-06-03T00:00:00.000Z","submittedOnDailyAt":"2026-06-10T00:00:00.000Z","title":"What Should Agents Say? Action-state Communication for Efficient Multi-Agent Systems","submittedOnDailyBy":{"_id":"65d7b983baa72790a1151923","avatarUrl":"/avatars/938531e84ca01a0c5a2a174057e3e9c5.svg","isPro":false,"fullname":"Chen Huang","user":"Albus-Chen","type":"user","name":"Albus-Chen"},"summary":"Multi-agent systems (MAS) built on large language models are typically organized around roles, pipelines, and turn schedules, while the content that agents pass to one another is often left as unconstrained natural language. However, this free-form communication can rapidly inflate token usage, consume the shared context window, and ultimately affect both system performance and inference cost. We analyze five common inter-agent communication strategies across two MAS topologies, finding that no fixed strategy is universally optimal. Instead, effective inter-agent messages consistently preserve action-centered information needed by downstream agents. Building on this, we propose the PACT (Protocolized Action-state Communication and Transmission), which treats inter-agent communication as a public state-update problem and projects each raw agent output into a compact action-state record before it enters shared history. Across different MAS topologies, PACT consistently improves the performance-cost trade-off, achieving comparable or stronger task performance with substantially fewer tokens. The gains extend to production coding harnesses: PACT lifts OpenHands' resolve rate at -10% tokens-per-resolved, and is resolve-neutral on SWE-agent while halving input tokens. Our code is publicly available at https://github.com/iNLP-Lab/PACT.","upvotes":4,"discussionId":"6a28c002e7d78ea7587e5299","githubRepo":"https://github.com/iNLP-Lab/PACT","githubRepoAddedBy":"user","ai_summary":"Multi-agent systems using large language models suffer from inefficient token consumption in agent-to-agent communication, which PACT addresses by structuring messages as compact action-state records that improve performance-cost trade-offs across different system architectures.","ai_keywords":["multi-agent systems","large language models","inter-agent communication","token usage","context window","action-centered information","protocolized action-state communication","public state-update problem","action-state record"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":4,"organization":{"_id":"68b82daee976083ccd80824b","name":"iNLP-Lab","fullname":"iNLP Lab @ SUTD","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/60dff6ae19a362a8c27862aa/3Ukf0b4f546tJM84zynTz.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"65d7b983baa72790a1151923","avatarUrl":"/avatars/938531e84ca01a0c5a2a174057e3e9c5.svg","isPro":false,"fullname":"Chen Huang","user":"Albus-Chen","type":"user"},{"_id":"667538975f7d5c8af7279078","avatarUrl":"/avatars/e8ad86fc75d885130efb2d5206fa5a8c.svg","isPro":false,"fullname":"Ryner Tan","user":"rryner","type":"user"},{"_id":"6316e5836d322c52490e7a5f","avatarUrl":"/avatars/c5919a3d9acff1a4d0b1219bb96a8fc4.svg","isPro":false,"fullname":"FengYi","user":"FMiMiY","type":"user"},{"_id":"6270324ebecab9e2dcf245de","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6270324ebecab9e2dcf245de/cMbtWSasyNlYc9hvsEEzt.jpeg","isPro":false,"fullname":"Kye Gomez","user":"kye","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"68b82daee976083ccd80824b","name":"iNLP-Lab","fullname":"iNLP Lab @ SUTD","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/60dff6ae19a362a8c27862aa/3Ukf0b4f546tJM84zynTz.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.05304.md"}">
Papers
arxiv:2606.05304

What Should Agents Say? Action-state Communication for Efficient Multi-Agent Systems

Published on Jun 3
· Submitted by
Chen Huang
on Jun 10
Authors:
,
,

Abstract

Multi-agent systems using large language models suffer from inefficient token consumption in agent-to-agent communication, which PACT addresses by structuring messages as compact action-state records that improve performance-cost trade-offs across different system architectures.

Multi-agent systems (MAS) built on large language models are typically organized around roles, pipelines, and turn schedules, while the content that agents pass to one another is often left as unconstrained natural language. However, this free-form communication can rapidly inflate token usage, consume the shared context window, and ultimately affect both system performance and inference cost. We analyze five common inter-agent communication strategies across two MAS topologies, finding that no fixed strategy is universally optimal. Instead, effective inter-agent messages consistently preserve action-centered information needed by downstream agents. Building on this, we propose the PACT (Protocolized Action-state Communication and Transmission), which treats inter-agent communication as a public state-update problem and projects each raw agent output into a compact action-state record before it enters shared history. Across different MAS topologies, PACT consistently improves the performance-cost trade-off, achieving comparable or stronger task performance with substantially fewer tokens. The gains extend to production coding harnesses: PACT lifts OpenHands' resolve rate at -10% tokens-per-resolved, and is resolve-neutral on SWE-agent while halving input tokens. Our code is publicly available at https://github.com/iNLP-Lab/PACT.

Community

Paper submitter about 15 hours ago

Multi-agent systems (MAS) built on large language models are typically organized around roles, pipelines, and turn schedules, while the content that agents pass to one another is often left as unconstrained natural language. However, this free-form communication can rapidly inflate token usage, consume the shared context window, and ultimately affect both system performance and inference cost. We analyze five common inter-agent communication strategies across two MAS topologies, finding that no fixed strategy is universally optimal. Instead, effective inter-agent messages consistently preserve action-centered information needed by downstream agents. Building on this, we propose the PACT (Protocolized Action-state Communication and Transmission), which treats inter-agent communication as a public state-update problem and projects each raw agent output into a compact action-state record before it enters shared history. Across different MAS topologies, PACT consistently improves the performance-cost trade-off, achieving comparable or stronger task performance with substantially fewer tokens. The gains extend to production coding harnesses: PACT lifts OpenHands' resolve rate at -10% tokens-per-resolved, and is resolve-neutral on SWE-agent while halving input tokens. Our code is publicly available at https://github.com/iNLP-Lab/PACT .

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images

· Sign up or log in to comment

Get this paper in your agent:

hf papers read 2606.05304
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2606.05304 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2606.05304 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2606.05304 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.

Discussion (0)

Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.

Sign in →

No comments yet. Sign in and be the first to say something.

More from Hugging Face Daily Papers