Emerging from Ground: Addressing Intent Deviation in Tool-Using Agents via Deriving Real Calls into Virtual Trajectories

Fuente: arXiv

Gespeichert in:

Bibliographische Detailangaben
Hauptverfasser:	Xiong, Qian, Huang, Yuekai, Yang, Bo, Zheng, Yujia, Li, Tianhao, Jiang, Ziyou, Chang, Zhiyuan, Li, Zhaoyang, Feng, Huanxiang, Li, Mingyang
Format:	Preprint
Veröffentlicht:	2026
Schlagworte:	Artificial Intelligence
Online-Zugang:	Acceder al recurso
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Ähnliche Einträge

Butterfly Effects in Toolchains: A Comprehensive Analysis of Failed Parameter Filling in LLM Tool-Agent Systems
von: Xiong, Qian, et al.
Veröffentlicht: (2025)

Know Thy Enemy: Securing LLMs Against Prompt Injection via Diverse Data Synthesis and Instruction-Level Chain-of-Thought Learning
von: Chang, Zhiyuan, et al.
Veröffentlicht: (2026)

Mimicking the Familiar: Dynamic Command Generation for Information Theft Attacks in LLM Tool-Learning System
von: Jiang, Ziyou, et al.
Veröffentlicht: (2025)

All Changes May Have Invariant Principles: Improving Ever-Shifting Harmful Meme Detection via Design Concept Reproduction
von: Jiang, Ziyou, et al.
Veröffentlicht: (2026)

Are All Prompt Components Value-Neutral? Understanding the Heterogeneous Adversarial Robustness of Dissected Prompt in Large Language Models
von: Zheng, Yujia, et al.
Veröffentlicht: (2025)

One Shot Dominance: Knowledge Poisoning Attack on Retrieval-Augmented Generation Systems
von: Chang, Zhiyuan, et al.
Veröffentlicht: (2025)

From Allies to Adversaries: Manipulating LLM Tool-Calling through Adversarial Injection
von: Wang, Haowei, et al.
Veröffentlicht: (2024)

Fall into a Pit, Gain in a Wit: Cognitive-Guided Harmful Meme Detection via Misjudgment Risk Pattern Retrieval
von: Wang, Wenshuo, et al.
Veröffentlicht: (2025)

Trajectory2Task: Training Robust Tool-Calling Agents with Synthesized Yet Verifiable Data for Complex User Intents
von: Wang, Ziyi, et al.
Veröffentlicht: (2026)

IntentMiner: Intent Inversion Attack via Tool Call Analysis in the Model Context Protocol
von: Yao, Yunhao, et al.
Veröffentlicht: (2025)

P‐4.35: Application of Virtual Reality Technology in the Diagnosis and Treatment of Psychological Disorders: An Electroencephalography (EEG)‐Based Approach
von: Weizhi Ma, et al.
Veröffentlicht: (2024)

Plant‐Derived Extracellular Vesicles as Potential Emerging Tools for Cancer Therapeutics
von: Lin Ding, et al.
Veröffentlicht: (2024)

What External Knowledge is Preferred by LLMs? Characterizing and Exploring Chain of Evidence in Imperfect Context for Multi-Hop QA
von: Chang, Zhiyuan, et al.
Veröffentlicht: (2024)

FinTrace: Holistic Trajectory-Level Evaluation of LLM Tool Calling for Long-Horizon Financial Tasks
von: Cao, Yupeng, et al.
Veröffentlicht: (2026)

Adversarial Testing for Visual Grounding via Image-Aware Property Reduction
von: Chang, Zhiyuan, et al.
Veröffentlicht: (2024)

P‐4.34: Virtual Reality Content Generation for Anxiety and Stress Management: Current Status, Challenges, and Future Developments
von: Yujia Zheng, et al.
Veröffentlicht: (2024)

SaFeR-ToolKit: Structured Reasoning via Virtual Tool Calling for Multimodal Safety
von: Xu, Zixuan, et al.
Veröffentlicht: (2026)

The Landscape of Emerging AI Agent Architectures for Reasoning, Planning, and Tool Calling: A Survey
von: Masterman, Tula, et al.
Veröffentlicht: (2024)

TrajDeleter: Enabling Trajectory Forgetting in Offline Reinforcement Learning Agents
von: Gong, Chen, et al.
Veröffentlicht: (2024)

Grounding Agent Memory in Contextual Intent
von: Yang, Ruozhen, et al.
Veröffentlicht: (2026)

The Evolution of Tool Use in LLM Agents: From Single-Tool Call to Multi-Tool Orchestration
von: Xu, Haoyuan, et al.
Veröffentlicht: (2026)

Agentic Link Construction for Environment and Intent Aware 6G Communication
von: Li, Zhaoyang, et al.
Veröffentlicht: (2025)

AEGIS: No Tool Call Left Unchecked -- A Pre-Execution Firewall and Audit Layer for AI Agents
von: Yuan, Aojie, et al.
Veröffentlicht: (2026)

VulRTex: A Reasoning-Guided Approach to Identify Vulnerabilities from Rich-Text Issue Report
von: Jiang, Ziyou, et al.
Veröffentlicht: (2025)

Efficient Tool-Calling Multi-Expert NPC Agent for Commonsense Persona-Grounded Dialogue
von: Nuriyev, Mahammad
Veröffentlicht: (2025)

W&D:Scaling Parallel Tool Calling for Efficient Deep Research Agents
von: Lin, Xiaoqiang, et al.
Veröffentlicht: (2026)

ShoppingBench: A Real-World Intent-Grounded Shopping Benchmark for LLM-based Agents
von: Wang, Jiangyuan, et al.
Veröffentlicht: (2025)

DroidCall: A Dataset for LLM-powered Android Intent Invocation
von: Xie, Weikai, et al.
Veröffentlicht: (2024)

VEglue: Testing Visual Entailment Systems via Object-Aligned Joint Erasing
von: Chang, Zhiyuan, et al.
Veröffentlicht: (2024)

Clawdrain: Exploiting Tool-Calling Chains for Stealthy Token Exhaustion in OpenClaw Agents
von: Dong, Ben, et al.
Veröffentlicht: (2026)

Computing Equilibrium beyond Unilateral Deviation
von: Liu, Mingyang, et al.
Veröffentlicht: (2026)

Agentic Search in the Wild: Intents and Trajectory Dynamics from 14M+ Real Search Requests
von: Ning, Jingjie, et al.
Veröffentlicht: (2026)

Speculative Interaction Agents: Building Real-Time Agents with Asynchronous I/O and Speculative Tool Calling
von: Hooper, Coleman, et al.
Veröffentlicht: (2026)

CDKFormer: Contextual Deviation Knowledge-Based Transformer for Long-Tail Trajectory Prediction
von: Lian, Yuansheng, et al.
Veröffentlicht: (2025)

Are Agents Probabilistic Automata? A Trace-Based, Memory-Constrained Theory of Agentic AI
von: Koohestani, Roham, et al.
Veröffentlicht: (2025)

Gecko: A Simulation Environment with Stateful Feedback for Refining Agent Tool Calls
von: Zhang, Zeyu, et al.
Veröffentlicht: (2026)

AgentStudio: A Toolkit for Building General Virtual Agents
von: Zheng, Longtao, et al.
Veröffentlicht: (2024)

Verification-Guided Context Optimization for Tool Calling via Hierarchical LLMs-as-Editors
von: Li, Henger, et al.
Veröffentlicht: (2025)

When Does Memory Help Multi-Trajectory Inference for Tool-Use LLM Agents?
von: Li, Xinzhe, et al.
Veröffentlicht: (2026)

Data-Enabled Neighboring Extremal: Case Study on Model-Free Trajectory Tracking for Robotic Arm
von: Vahidi-Moghaddam, Amin, et al.
Veröffentlicht: (2025)

if (filemtime($filename) + $maxlifetime < time()) { unlink($filename); $count++; } } return $count; } /** * A function that is called internally when session data is to be saved. * * @param string $sessId The current session ID * @param string $data The session data to write * * @return bool */ protected function saveSession($sessId, $data): bool { $sessFile = $this->path . '/sess_' . $sessId; if ($handle = fopen($sessFile, 'w')) { $return = false; // Lock the file for exclusive access to avoid issues with multiple // processes writing session simultaneously: if (flock($handle, LOCK_EX)) { $return = fwrite($handle, $data); // Make sure that there's no trailing data by truncating the file to // the correct length: ftruncate($handle, strlen($data)); fflush($handle); flock($handle, LOCK_UN); } fclose($handle); if ($return !== false) { return true; } } // If we got this far, something went wrong with the file output... // It is tempting to throw an exception here, but this code is called // outside of the context of exception handling, so all we can do is // echo a message.

// Anecdotal testing Today and Yesterday seems to indicate destroy() // is called by the garbage collector and everything is good. // Something to keep in mind though. return 0; } /** * Write function that is called when session data is to be saved. * * @param string $sessId The current session ID * @param string $data The session data to write * * @return bool */ public function write($sessId, $data): bool { if ($this->writesDisabled) { return true; } return $this->saveSession($sessId, $data); } /** * A function that is called internally when session data is to be saved. * * @param string $sessId The current session ID * @param string $data The session data to write * * @return bool */ abstract protected function saveSession($sessId, $data): bool; }

* * @return void */ public function writeClose() { // The assumption is that we're using PHP's ext/session. // session_write_close() will actually overwrite $_SESSION with an // empty array on completion -- which leads to a mismatch between what // is in the storage object and $_SESSION. To get around this, we // temporarily reset $_SESSION to an array, and then re-link it to // the storage object. // // Additionally, while you _can_ write to $_SESSION following a // session_write_close() operation, no changes made to it will be // flushed to the session handler. As such, we now mark the storage // object isImmutable. $storage = $this->getStorage(); if (! $storage->isImmutable()) { $_SESSION = $storage->toArray(true); session_write_close(); $storage->fromArray($_SESSION); $storage->markImmutable(); } } /** * Attempt to set the session name * * If the session has already been started, or if the name provided fails * validation, an exception will be raised. * * @param string $name * @return SessionManager * @throws Exception\InvalidArgumentException */ public function setName($name) { if ($this->sessionExists()) { throw new Exception\InvalidArgumentException( 'Cannot set session name after a session has already started'

Environment & details:

GET Data empty

POST Data empty

Files empty

Key	Value
ui	"standard"
language	"de"
VUFIND_SESSION	"0418e18c12839c66ebf0fc6ef5664a24"

Session

Key	Value
__Laminas	array:2 [`"_REQUEST_ACCESS_TIME" => 1790823744.8332 "_VALID" => array:1 [ "Laminas\Session\Validator\Id" => "0418e18c12839c66ebf0fc6ef5664a24" ]`]
SessionState	Laminas\Stdlib\ArrayObject {#938}
FlashMessenger	Laminas\Stdlib\ArrayObject {#1433}
SessionHelper	Laminas\Stdlib\ArrayObject {#1759}

Server/Request Data

Key	Value
REDIRECT_VUFIND_LOCAL_DIR	"/usr/local/vufind/local"
REDIRECT_HTTPS	"on"
REDIRECT_SSL_TLS_SNI	"info.umar.mx"
REDIRECT_STATUS	"200"
VUFIND_LOCAL_DIR	"/usr/local/vufind/local"
HTTPS	"on"
SSL_TLS_SNI	"info.umar.mx"
HTTP_ACCEPT	"/"
HTTP_USER_AGENT	"Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; ClaudeBot/1.0; +claudebot@anthropic.com)"
HTTP_COOKIE	"ui=standard; language=de; VUFIND_SESSION=0418e18c12839c66ebf0fc6ef5664a24"
HTTP_ACCEPT_ENCODING	"gzip, br, zstd, deflate"
HTTP_HOST	"info.umar.mx"
PATH	"/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/snap/bin"
SERVER_SIGNATURE	"<address>Apache/2.4.66 (Ubuntu) Server at info.umar.mx Port 443</address>\n"
SERVER_SOFTWARE	"Apache/2.4.66 (Ubuntu)"
SERVER_NAME	"info.umar.mx"
SERVER_ADDR	"200.23.222.19"
SERVER_PORT	"443"
REMOTE_ADDR	"216.73.216.147"
DOCUMENT_ROOT	"/usr/local/vufind/public"
REQUEST_SCHEME	"https"
CONTEXT_PREFIX	""
CONTEXT_DOCUMENT_ROOT	"/usr/local/vufind/public"
SERVER_ADMIN	"admin@localhost"
SCRIPT_FILENAME	"/usr/local/vufind/public/index.php"
REMOTE_PORT	"8499"
REDIRECT_URL	""" /Record/\n arxiv_https___arxiv_org_abs_2601_15120/Similar """
GATEWAY_INTERFACE	"CGI/1.1"
SERVER_PROTOCOL	"HTTP/1.1"
REQUEST_METHOD	"GET"
QUERY_STRING	""
REQUEST_URI	"/Record/%0Aarxiv_https___arxiv_org_abs_2601_15120/Similar"
SCRIPT_NAME	"/index.php"
PHP_SELF	"/index.php"
REQUEST_TIME_FLOAT	1790823744.8211
REQUEST_TIME	1790823744

Environment Variables empty

Registered Handlers

0. Whoops\Handler\PrettyPageHandler