Hugging Face Daily Papers · August 18, 2026 · 4 min read

Agentic Transaction: Towards ACID-Compliant Agent Systems

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

Like Read original ↗

Can LLM agents have transactions like databases?\n• Agentic Transaction — a transactional abstraction for reliable long-horizon LLM agent execution. • ACID for Agents — Semantic Atomicity, Consistency, Isolation, and Durability. • ACID-Agent — a concrete data-agent system that realizes these guarantees in real-world data analysis. • +10.6% over Claude Code on KramaBench.\n<a href=\"https://cdn-uploads.huggingface.co/production/uploads/654cdf9895edd6dffccdcf65/8rd8zfkoGUmw76QNimjaY.png\" rel=\"nofollow\"><img src=\"https://cdn-uploads.huggingface.co/production/uploads/654cdf9895edd6dffccdcf65/8rd8zfkoGUmw76QNimjaY.png\" alt=\"example\"></a>\n","updatedAt":"2026-08-18T02:01:25.078Z","author":{"_id":"654cdf9895edd6dffccdcf65","avatarUrl":"/avatars/1e3ae4c096092388b2c7d0bc42869c9d.svg","fullname":"Zhaoyan Sun","name":"curtis-sun","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":1,"identifiedLanguage":{"language":"en","probability":0.8666020631790161},"editors":["curtis-sun"],"editorAvatarUrls":["/avatars/1e3ae4c096092388b2c7d0bc42869c9d.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.13900","authors":[{"_id":"6a83bac8675db694db8cd446","name":"Zhaoyan Sun","hidden":false},{"_id":"6a83bac8675db694db8cd447","name":"Xiaoxiao Wang","hidden":false},{"_id":"6a83bac8675db694db8cd448","name":"Guoliang Li","hidden":false}],"publishedAt":"2026-08-14T00:00:00.000Z","submittedOnDailyAt":"2026-08-18T00:00:00.000Z","title":"Agentic Transaction: Towards ACID-Compliant Agent Systems","submittedOnDailyBy":{"_id":"654cdf9895edd6dffccdcf65","avatarUrl":"/avatars/1e3ae4c096092388b2c7d0bc42869c9d.svg","isPro":false,"fullname":"Zhaoyan Sun","user":"curtis-sun","type":"user","name":"curtis-sun"},"summary":"Large language model (LLM) agents are evolving from conversational assistants into autonomous systems that execute long-horizon tasks through reasoning, tool use, code generation, and workspace manipulation. As agents increasingly operate over persistent environments and multi-step workflows, they face challenges analogous to those addressed by transactional database systems: reliable execution, consistent outcomes, safe concurrency, and durable state management. We introduce the concept of an agentic transaction and propose an ACID-compliant agent system framework that reinterprets the classical ACID properties for agent execution through four semantic guarantees: Semantic Atomicity, Semantic Consistency, Semantic Isolation, and Semantic Durability. Together, these properties provide a principled foundation for building reliable agent systems despite model uncertainty and dynamic execution environments. To instantiate this framework, we develop an ACID-compliant data agent that realizes these guarantees through transactional exploration-execution-validation cycles, transactional skill hubs, confidence divergence-based validation, semantic dependency-aware isolation, and transaction-aware semantic state management. Experimental results on widely used benchmarks show that our system achieves a 10.6% improvement over state-of-the-art agents, including Claude Code. This work opens a broader research agenda on extending transactional principles and system architectures toward building trustworthy, scalable, and self-evolving AI agent systems.","upvotes":20,"discussionId":"6a83bac9675db694db8cd449","githubRepo":"https://github.com/TsinghuaDatabaseGroup/ACID-Agent","githubRepoAddedBy":"user","ai_summary":"An ACID-compliant framework for agentic transactions introduces semantic guarantees to ensure reliable, isolated, and durable execution of long-horizon LLM agent workflows.","ai_keywords":["agentic transaction","ACID-compliant agent system","Semantic Atomicity","Semantic Consistency","Semantic Isolation","Semantic Durability","transactional exploration-execution-validation","transactional skill hubs","confidence divergence-based validation","semantic dependency-aware isolation","transaction-aware semantic state management"],"ai_summary_model":"thinkingmachines/Inkling-Small","githubStars":3,"organization":{"_id":"628735cbc83a2d6ab8d14a66","name":"Tsinghua","fullname":"Tsinghua University","avatar":"https://www.gravatar.com/avatar/6c5c1441e3283e7543342e59277ea219?d=retro&size=100"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"654cdf9895edd6dffccdcf65","avatarUrl":"/avatars/1e3ae4c096092388b2c7d0bc42869c9d.svg","isPro":false,"fullname":"Zhaoyan Sun","user":"curtis-sun","type":"user"},{"_id":"67f5d9f4f7a398c2a27166d3","avatarUrl":"/avatars/4d339fe7b9b760c2f936823f6de9eab2.svg","isPro":false,"fullname":"KANAME YOKOYAMA","user":"KanameYOkoYAMA","type":"user"},{"_id":"68308fb0cddb5ce16ae19445","avatarUrl":"/avatars/a5faa2cb600176a0050e06bb98193558.svg","isPro":false,"fullname":"Hao","user":"polar1ss","type":"user"},{"_id":"67baf875419becb9ac7c9aef","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/GEsp0WwhMg1oCPGWIx7zW.png","isPro":false,"fullname":"Ji Sun","user":"greatjiTHU","type":"user"},{"_id":"63ea1e3ddedfeebe54cff022","avatarUrl":"/avatars/02290409a94f89aab0ae474c5424d53d.svg","isPro":false,"fullname":"Yonghao Song","user":"eeyhsong","type":"user"},{"_id":"66c09891e61ccd71d7660447","avatarUrl":"/avatars/890afa8794737fd8452a9fb60348cfd9.svg","isPro":false,"fullname":"Yuanli","user":"Macly","type":"user"},{"_id":"6451275d6db3be7ee57a7108","avatarUrl":"/avatars/a77d8732b7a8413b84ca4b1f04a0ab71.svg","isPro":false,"fullname":"HD","user":"Dtech21","type":"user"},{"_id":"63ac5701c21e60a3e9b58aa7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63ac5701c21e60a3e9b58aa7/g6EX7diOpuA94R2ab-rZC.png","isPro":true,"fullname":"Dipankar Sarkar","user":"dipankarsarkar","type":"user"},{"_id":"67be8bfad1e0f62eb8edba08","avatarUrl":"/avatars/c5c046fa4f87a630382eb575a6f3efd5.svg","isPro":false,"fullname":"jiaqi Tian","user":"jiaqiTian","type":"user"},{"_id":"6463554dd2044cd1d7c6e0bf","avatarUrl":"/avatars/d7653623117268c545a7063fec69664b.svg","isPro":false,"fullname":"Bingzheng Wei","user":"Bingzheng","type":"user"},{"_id":"6874c35fb3a50b6735520dc1","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/R5XW2PoiKbjHtDPbtQLDc.png","isPro":false,"fullname":"Sophia Cai","user":"IneffableCPP","type":"user"},{"_id":"67be87b95a29257e4850edb8","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/RLrVf_p1KdbIrlcyUp31O.png","isPro":false,"fullname":"Yuwei Huang","user":"hyw498169842","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"628735cbc83a2d6ab8d14a66","name":"Tsinghua","fullname":"Tsinghua University","avatar":"https://www.gravatar.com/avatar/6c5c1441e3283e7543342e59277ea219?d=retro&size=100"},"query":{}}">

Papers

arxiv:2608.13900

Agentic Transaction: Towards ACID-Compliant Agent Systems

Published on Aug 14

· Submitted by

Zhaoyan Sun on Aug 18

Tsinghua University

Upvote

Authors:

Abstract

An ACID-compliant framework for agentic transactions introduces semantic guarantees to ensure reliable, isolated, and durable execution of long-horizon LLM agent workflows.

Generated by thinkingmachines/Inkling-Small

Large language model (LLM) agents are evolving from conversational assistants into autonomous systems that execute long-horizon tasks through reasoning, tool use, code generation, and workspace manipulation. As agents increasingly operate over persistent environments and multi-step workflows, they face challenges analogous to those addressed by transactional database systems: reliable execution, consistent outcomes, safe concurrency, and durable state management. We introduce the concept of an agentic transaction and propose an ACID-compliant agent system framework that reinterprets the classical ACID properties for agent execution through four semantic guarantees: Semantic Atomicity, Semantic Consistency, Semantic Isolation, and Semantic Durability. Together, these properties provide a principled foundation for building reliable agent systems despite model uncertainty and dynamic execution environments. To instantiate this framework, we develop an ACID-compliant data agent that realizes these guarantees through transactional exploration-execution-validation cycles, transactional skill hubs, confidence divergence-based validation, semantic dependency-aware isolation, and transaction-aware semantic state management. Experimental results on widely used benchmarks show that our system achieves a 10.6% improvement over state-of-the-art agents, including Claude Code. This work opens a broader research agenda on extending transactional principles and system architectures toward building trustworthy, scalable, and self-evolving AI agent systems.

View arXiv page View PDF GitHub 3 Add to collection

Community

curtis-sun

Paper submitter about 7 hours ago edited about 7 hours ago

Can LLM agents have transactions like databases?

• Agentic Transaction — a transactional abstraction for reliable long-horizon LLM agent execution.
• ACID for Agents — Semantic Atomicity, Consistency, Isolation, and Durability.
• ACID-Agent — a concrete data-agent system that realizes these guarantees in real-world data analysis.
• +10.6% over Claude Code on KramaBench.