MARC v1: An Open-Source Multi-Agent Framework for Clinical AI Reasoning and Coordination
- ID: 378f9f88
- 原文链接: https://arxiv.org/abs/2608.13476
- PDF: https://arxiv.org/pdf/2608.13476v1
- 作者: Saisha Shetty, Satvik Tripathi, Austin Lin, Colin Zhao, Theodore Kim, Don Enwerem, Jacinta Arnold, Shahriar Faghani, Tessa S Cook
- 日期: 2026-08-13
- 更新: N/A
- 分类: agents
- 来源类型: paper
- 标签: multi-agent, clinical-ai, orchestration, open-source, arxiv
- 质量评分: 4/5
- 抓取时间: 2026-08-17T04:23:51+00:00Z
中文导读
MARC（Multi-Agent Reasoning and Coordination）是面向临床推理的开源多智能体框架：用确定性编排替代单体 LLM 提示，抽取推理答案生成评估由角色化 agent 分工，上下文显式传递中间产物可追溯，支持阶段级失败归因配套 Decomposer 模块可从自然语言任务描述自动生成各 agent 专用提示，免去手工 prompt 工程；全程 YAML 配置无需改代码，支持 API 与本地 CPU 两种部署，面向无编程背景的临床领域专家
为什么值得关注
MARC（Multi-Agent Reasoning and Coordination）是面向临床推理的开源多智能体框架：用确定性编排替代单体 LLM 提示，抽取推理答案生成评估由角色化 agent 分工，上下文显式传递中间产物可追溯。
收录理由：把可归因的多 agent 编排做成临床可落地的开源实现，阶段级失败定位思路可迁移到其他专业领域
关键信息
- 论文标题：MARC v1: An Open-Source Multi-Agent Framework for Clinical AI Reasoning and Coordination
- 作者：Saisha Shetty, Satvik Tripathi, Austin Lin, Colin Zhao, Theodore Kim, Don Enwerem, Jacinta Arnold, Shahriar Faghani, Tessa S Cook
- arXiv：https://arxiv.org/abs/2608.13476
- 发布时间：2026-08-13
- arXiv 分类：cs.AI, cs.CL
- 关联标签：multi-agent, clinical-ai, orchestration, open-source, arxiv
English Abstract
We present Multi-Agent Reasoning and Coordination (MARC), an open-source framework that replaces monolithic LLM prompting with deterministic multi-agent orchestration for clinical reasoning. MARC coordinates role-specialized agents for extraction, reasoning, answer generation, and evaluation, with explicit context passing and traceable intermediate outputs, enabling stage-wise failure attribution. We additionally introduce a Decomposer module that generates task-specific agent prompts from a plain-language description, eliminating manual prompt engineering. The framework supports both API-based and local CPU-compatible deployments and is entirely configurable via YAML, without code modifications. MARC is designed to be model-agnostic, interpretable, and accessible to clinical domain experts without programming expertise. The full framework is available at https://github.com/Penn-RAIL/MARC-v1.
English Summary
MARC is an open-source framework that replaces monolithic LLM prompting with deterministic multi-agent orchestration for clinical reasoning. Role-specialized agents handle extraction, reasoning, answer generation, and evaluation with explicit context passing and traceable intermediate outputs, enabling stage-wise failure attribution. A Decomposer module generates task-specific agent prompts from a plain-language description, eliminating manual prompt engineering. Deployment supports API-based and local CPU-compatible setups, fully configurable via YAML without code changes; the framework is model-agnostic and aimed at clinical domain experts. Code: github.com/Penn-RAIL/MARC-v1.
Obsidian Notes
- 内容由
opencli arxiv paper拉取 arXiv 元数据与摘要生成。 - 中文导读与价值判断锚定于条目与论文摘要，未补充摘要之外的内容。