ASA: Training-Free Representation Engineering for Tool-Calling Agents

• URL: http://arxiv.org/abs/2602.04935v2
• Date: Sun, 08 Feb 2026 09:26:59 GMT
• Title: ASA: Training-Free Representation Engineering for Tool-Calling Agents
• Abstract summary: Activation Steering Adapter (ASA) is a training-free, inference-time controller that performs a single-shot mid-layer intervention and targets tool domains.<n>On MTU-Bench with Qwen2.5-1.5B, ASA improves strict tool-use F1 from 0.18 to 0.50 while reducing the false positive rate from 0.15 to 0.05, using only about 20KB of portable assets and no weight updates.
• Score: 27.41214164139876
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Adapting LLM agents to domain-specific tool calling remains notably brittle under evolving interfaces. Prompt and schema engineering is easy to deploy but often fragile under distribution shift and strict parsers, while continual parameter-efficient fine-tuning improves reliability at the cost of training, maintenance, and potential forgetting. We identify a critical Lazy Agent failure mode where tool necessity is nearly perfectly decodable from mid-layer activations, yet the model remains conservative in entering tool mode, revealing a representation-behavior gap. We propose Activation Steering Adapter (ASA), a training-free, inference-time controller that performs a single-shot mid-layer intervention and targets tool domains via a router-conditioned mixture of steering vectors with a probe-guided signed gate to amplify true intent while suppressing spurious triggers. On MTU-Bench with Qwen2.5-1.5B, ASA improves strict tool-use F1 from 0.18 to 0.50 while reducing the false positive rate from 0.15 to 0.05, using only about 20KB of portable assets and no weight updates.

Related papers

• Forgotten in Weights, Recovered by Tools: Agentic Tool Unlearning for LLM Agents [54.050506620820784]
  Agentic Tool Unlearning (ATU) aims to reduce both parametric recall and tool-mediated recovery while preserving normal tool use for retained knowledge.<n>ATU achieves a better balance between target forgetting and retained utility, making unlearning more robust under tool-augmented agent deployment.
  arXiv  Detail & Related papers  (2026-08-21T18:34:48Z)
• ParaVT: Taming the Tool Prior Paradox for Parallel Tool Use in Agentic Video Reinforcement Learning [91.51460129144654]
  We introduce ParaVT, the first multi-agent end-to-end RL-trained framework for Parallel Video Tool calling.<n>We show that ParaVT improves over the Qwen3-VL baseline by +7.9% on average, with PARA-GRPO lifting training-time format compliance from 0.13 to 0.64.
  arXiv  Detail & Related papers  (2026-05-19T18:01:26Z)
• Beyond the Black Box: Interpretability of Agentic AI Tool Use [0.0]
  We introduce a mechanistic-interpretability toolkit built on Sparse Autoencoders and linear probes.<n>The framework reads model states before each action and infers both whether a tool is needed and how consequential the next tool action is likely to be.<n>We train the probes on multi-step trajectories from the NVIDIA Nemotron function-calling dataset and apply the same workflow to GPT-OSS 20B and Gemma 3 27B models.
  arXiv  Detail & Related papers  (2026-05-07T19:47:30Z)
• Waking Up Blind: Cold-Start Optimization of Supervision-Free Agentic Trajectories for Grounded Visual Perception [18.565166776233223]
  Small Vision-Language Models (SVLMs) are efficient task controllers but often suffer from visual brittleness and poor tool orchestration.<n>We propose Self-supervised Perception Enabled by Cascaded Tool Rollout Alignment (SPECTRA)<n>SPECTRA bootstraps agentic capabilities via Coldstart Reinforcement Learning for SVLMs.
  arXiv  Detail & Related papers  (2026-04-19T15:06:30Z)
• Controllable and Verifiable Tool-Use Data Synthesis for Agentic Reinforcement Learning [82.89535601592739]
  We propose a two-stage pipeline that generates reliable base tool-use trajectories through self-evolving synthesis with multi-level validation.<n>These augmentations introduce distractor tools, indirect or ambiguous user queries, and noisy, multi-format, or erroneous tool outputs.<n>This design enables automatic reward computation via reference matching for standard cases and lightweight judge-assisted verification for special behaviors such as error detection.
  arXiv  Detail & Related papers  (2026-04-10T18:38:52Z)
• AgentProcessBench: Diagnosing Step-Level Process Quality in Tool-Using Agents [50.481033105867205]
  We introduce AgentProcessBench, the first benchmark dedicated to evaluating step-level effectiveness in realistic, tool-augmented trajectories.<n>The benchmark comprises 1,000 diverse trajectories and 8,509 human-labeled step annotations with 89.1% inter-annotator agreement.<n>It features a ternary labeling scheme to capture exploration and an error propagation rule to reduce labeling ambiguity.
  arXiv  Detail & Related papers  (2026-03-15T16:13:58Z)
• Unsafer in Many Turns: Benchmarking and Defending Multi-Turn Safety Risks in Tool-Using Agents [68.20752678837377]
  We propose a principled taxonomy that transforms single-turn harmful tasks into multi-turn attack sequences.<n>Using this taxonomy, we construct MT-AgentRisk, the first benchmark to evaluate multi-turn tool-using agent safety.<n>We propose ToolShield, a training-free, tool-agnostic, self-exploration defense.
  arXiv  Detail & Related papers  (2026-02-13T18:38:18Z)
• Sponge Tool Attack: Stealthy Denial-of-Efficiency against Tool-Augmented Agentic Reasoning [58.432996881401415]
  Recent work augments large language models (LLMs) with external tools to enable agentic reasoning.<n>We propose Sponge Tool Attack (STA), which disrupts agentic reasoning solely by rewriting the input prompt.<n>STA generates benign-looking prompt rewrites from the original one with high semantic fidelity.
  arXiv  Detail & Related papers  (2026-01-24T19:36:51Z)
• Beyond Max Tokens: Stealthy Resource Amplification via Tool Calling Chains in LLM Agents [31.789859492703016]
  The agent-tool communication loop is a critical attack surface in Large Language Model (LLM) agents.<n>Existing Denial-of-Service (DoS) attacks are ineffective for this new paradigm.<n>We introduce a stealthy, multi-turn economic DoS attack that operates at the tool layer under the guise of a correctly completed task.
  arXiv  Detail & Related papers  (2026-01-16T02:47:45Z)
• RISER: Orchestrating Latent Reasoning Skills for Adaptive Activation Steering [62.63376387138257]
  We propose a plug-and-play intervention framework that adaptively steers large language models (LLMs) reasoning in activation space.<n>RISER constructs a library of reusable reasoning vectors and employs a lightweight Router to dynamically compose them for each input.<n>The Router is optimized via reinforcement learning under task-level rewards, activating latent cognitive primitives in an emergent and compositional manner.
  arXiv  Detail & Related papers  (2026-01-14T08:04:33Z)
• ET-Agent: Incentivizing Effective Tool-Integrated Reasoning Agent via Behavior Calibration [68.89572566071575]
  ETAgent is a training framework for calibrating agent's tool-use behavior.<n>It is designed to progressively calibrate erroneous behavioral patterns to optimal behaviors.
  arXiv  Detail & Related papers  (2026-01-11T11:05:26Z)
• PAGen: Phase-guided Amplitude Generation for Domain-adaptive Object Detection [15.55359477953804]
  Unsupervised domain adaptation (UDA) greatly facilitates the deployment of neural networks across diverse environments.<n>We present a simple yet effective UDA method that learns to adapt image styles in the frequency domain to reduce the discrepancy between source and target domains.
  arXiv  Detail & Related papers  (2025-11-27T02:22:37Z)
• UltraCUA: A Foundation Model for Computer Use Agents with Hybrid Action [77.63125913907771]
  We present UltraCUA, a foundation model that bridges the gap between GUI primitives and high-level programmatic tool calls.<n>Experiments with our 7B and 32B models demonstrate substantial improvements over state-of-the-art agents.
  arXiv  Detail & Related papers  (2025-10-20T17:48:26Z)
• Adaptive Minds: Empowering Agents with LoRA-as-Tools [0.0]
  Adaptive Minds is an agentic system that treats LoRA adapters as domain-specific tools.<n>System is built with LangGraph for workflow management, supports both API and web interfaces, and is fully open source.
  arXiv  Detail & Related papers  (2025-10-17T08:10:06Z)
• Tool-R1: Sample-Efficient Reinforcement Learning for Agentic Tool Use [50.02614257515131]
  Large language models (LLMs) have demonstrated strong capabilities in language understanding and reasoning.<n>We propose Tool-R1, a reinforcement learning framework that enables LLMs to perform general, compositional, and multi-step tool use.
  arXiv  Detail & Related papers  (2025-09-16T09:22:21Z)
• Generalized Adaptive Transfer Network: Enhancing Transfer Learning in Reinforcement Learning Across Domains [11.705324423141606]
  Transfer learning in Reinforcement Learning (RL) enables agents to leverage knowledge from source tasks to accelerate learning in target tasks.<n>This paper introduces the Generalized Adaptive Transfer Network (GATN), a deep RL architecture designed to tackle task generalization across domains.<n>GATN employs a domain-agnostic representation module, a robustness-aware policy adapter, and an efficient transfer scheduler to achieve these goals.
  arXiv  Detail & Related papers  (2025-07-02T21:33:48Z)
• Patterns and Mechanisms of Contrastive Activation Engineering [0.374490703387131]
  CAE has the potential to introduce a new paradigm of flexible, task-specific behavior tuning.<n>We analyze the performance of CAE in in in-distribution, out-of-distribution settings, evaluate drawbacks, and begin to develop comprehensive guidelines for its effective deployment.
  arXiv  Detail & Related papers  (2025-05-06T05:15:12Z)
• AegisLLM: Scaling Agentic Systems for Self-Reflective Defense in LLM Security [74.22452069013289]
  AegisLLM is a cooperative multi-agent defense against adversarial attacks and information leakage.<n>We show that scaling agentic reasoning system at test-time substantially enhances robustness without compromising model utility.<n> Comprehensive evaluations across key threat scenarios, including unlearning and jailbreaking, demonstrate the effectiveness of AegisLLM.
  arXiv  Detail & Related papers  (2025-04-29T17:36:05Z)
• PCA-Featured Transformer for Jamming Detection in 5G UAV Networks [0.5999777817331317]
  Unmanned Aerial Vehicles (UAVs) face significant security risks from jamming attacks, which can compromise network functionality.<n>Traditional detection methods often fall short when confronting AI-powered jamming that dynamically modifies its behavior.<n>We introduce a novel U-shaped transformer architecture to refine feature representations for improved wireless security.
  arXiv  Detail & Related papers  (2024-12-19T16:13:04Z)
• To Adapt or Not to Adapt? Real-Time Adaptation for Semantic Segmentation [45.74645069886128]
  HAMLET is a Hardware-Aware Modular Least Expensive Training framework for real-time domain adaptation. Our approach is capable of performing semantic segmentation while simultaneously adapting at more than 29FPS on a single consumer-grade GPU.
  arXiv  Detail & Related papers  (2023-07-27T17:59:59Z)
• Multi-Prompt Alignment for Multi-Source Unsupervised Domain Adaptation [86.02485817444216]
  We introduce Multi-Prompt Alignment (MPA), a simple yet efficient framework for multi-source UDA. MPA denoises the learned prompts through an auto-encoding process and aligns them by maximizing the agreement of all the reconstructed prompts. Experiments show that MPA achieves state-of-the-art results on three popular datasets with an impressive average accuracy of 54.1% on DomainNet.
  arXiv  Detail & Related papers  (2022-09-30T03:40:10Z)
• Stagewise Unsupervised Domain Adaptation with Adversarial Self-Training for Road Segmentation of Remote Sensing Images [93.50240389540252]
  Road segmentation from remote sensing images is a challenging task with wide ranges of application potentials. We propose a novel stagewise domain adaptation model called RoadDA to address the domain shift (DS) issue in this field. Experiment results on two benchmarks demonstrate that RoadDA can efficiently reduce the domain gap and outperforms state-of-the-art methods.
  arXiv  Detail & Related papers  (2021-08-28T09:29:14Z)
• Towards Safe Control of Continuum Manipulator Using Shielded Multiagent Reinforcement Learning [1.2647816797166165]
  The control of the robot is formulated as a one-DoF, one agent problem in the MADQN framework to improve the learning efficiency. Shielded MADQN enabled the robot to perform point and trajectory tracking with submillimeter root mean square errors under external loads.
  arXiv  Detail & Related papers  (2021-06-15T05:55:05Z)
• AFAN: Augmented Feature Alignment Network for Cross-Domain Object Detection [90.18752912204778]
  Unsupervised domain adaptation for object detection is a challenging problem with many real-world applications. We propose a novel augmented feature alignment network (AFAN) which integrates intermediate domain image generation and domain-adversarial training. Our approach significantly outperforms the state-of-the-art methods on standard benchmarks for both similar and dissimilar domain adaptations.
  arXiv  Detail & Related papers  (2021-06-10T05:01:20Z)
• DA-DETR: Domain Adaptive Detection Transformer with Information Fusion [53.25930448542148]
  DA-DETR is a domain adaptive object detection transformer that introduces information fusion for effective transfer from a labeled source domain to an unlabeled target domain. We introduce a novel CNN-Transformer Blender (CTBlender) that fuses the CNN features and Transformer features ingeniously for effective feature alignment and knowledge transfer across domains. CTBlender employs the Transformer features to modulate the CNN features across multiple scales where the high-level semantic information and the low-level spatial information are fused for accurate object identification and localization.
  arXiv  Detail & Related papers  (2021-03-31T13:55:56Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.