跳到正文
HuggingFace Daily Papers(社區熱門論文)·· 28 天前AI 評分50

MCP 與 CLI 語言智能體介面的受控研究:能力相同,行為各異

Matched Capabilities, Different Behaviors: A Controlled Study of MCP and CLI Interfaces for Language Agents

AI 導讀

一項受控研究比較 MCP 與 CLI 介面,發現即使可執行能力相同,介面呈現方式仍會改變智能體行為、任務表現及互動成本。六個模型在三個互動環境的 2,004 次測試中,有 94.3% 出現不同軌跡;DART 在六項評測設定中最多提升 9.0 分,並在三個環境中每項任務最多減少 27.1k 個 MCP token。

正文

Image 1: Hugging Face's logoHugging Face


Papers

arxiv:2610.34298

Matched Capabilities, Different Behaviors: A Controlled Study of MCP and CLI Interfaces for Language Agents

Published on Oct 25

·Submitted byImage 2 zhenning shion Oct 30

·Image 3: ModalityDanceModalityDance

- [x] Upvote 1

  • Image 4

Authors:

Image 5Zhenning Shi,

Liang Pang,

Yongqi Li

Abstract

Matched interface representations significantly alter agent behavior despite identical executable capabilities, and targeted repair of interface-specific realization improves performance while reducing cost.

Generated by thinkingmachines/Inkling-Small

Language agents increasingly act through external tools and interactive environments, making the interface between the model and its executable capabilities part of the decision process. Model Context Protocol (MCP) and command-line interfaces (CLI) can expose the same semantic operations through markedly different model-facing representations. We study whether these representational differences change agent behavior when executable capabilities are matched. We introduce Matched Interface Comparison and Analysis (MICA), which renders MCP and CLI from a shared semantic action contract and routes decoded actions through the same runtime. Across six models and three interactive environments, matched interfaces produce substantial differences in task performance and interaction cost. These effects extend beyond surface validity to semantic action selection, failure pathways, and recovery. Matched-state probes localize sensitivity to capability presentation, realization, and feedback. Paired natural trajectories reveal structured divergence in 94.3% of 2,004 episodes and show that early differences propagate through subsequent interaction. This analysis identifies interface realization as a direct intervention boundary. We therefore introduce Decoupled Action Realization and Targeted Repair (DART), which preserves validated semantic commitments while repairing interface-specific realization. DART improves benchmark score in all six evaluated settings by up to 9.0 points and reduces MCP token cost in all three environments by up to 27.1k tokens per task. Together, these results establish interface representation as part of the closed-loop decision environment shaping agent behavior.

View arXiv pageView PDFProject pageGitHub 0Add to collection

Community

Image 6

sznnzs

Paper author Paper submitter21 minutes ago

We introduce MICA, a controlled study of MCP and CLI interfaces with matched executable capabilities, spanning six models and three interactive environments. We find substantial differences in performance, interaction cost, and agent behavior.

We also introduce DART, which preserves validated semantic actions while repairing interface-specific realization. DART improves benchmark scores in all six evaluated settings by up to 9.0 points and reduces MCP token cost by up to 27.1k tokens per task.

Project page: https://modalitydance.github.io/MICA/

Code: https://github.com/ModalityDance/MICA

Reply

Edit Preview

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

Tap or paste here to upload images

Comment ·Sign up or log in to comment

- [x] Upvote 1

  • Image 7

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2610.34298 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/2610.34298 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/2610.34298 in a Space README.md to link it from this page.

Collections including this paper 1

MICA Collection Matched Capabilities, Different Behaviors: A Controlled Study of MCP and CLI Interfaces for Language Agents•1 item•Updated about 1 hour ago

System theme

Company

TOSPrivacyAboutCareers

Website

ModelsDatasetsSpacesPricingDocs

來源:HuggingFace Daily Papers(社區熱門論文) · huggingface.co