A Few Neurons Reveal When LLMs Misuse Tools: Sparse Detection and Selective Steering for Reliable Tool Use

• URL: http://arxiv.org/abs/2608.00218v1
• Date: Fri, 31 Jul 2026 19:03:44 GMT
• Title: A Few Neurons Reveal When LLMs Misuse Tools: Sparse Detection and Selective Steering for Reliable Tool Use
• Abstract summary: We introduce PRISMS (Probing Representations In Support of Monitoring and Steering), a closed-loop framework that shares a failure-specific neuron basis between sparse detection and activation.<n>Across six models, PRISMS reduces pooled over-calling rate by 80% while increasing tool-required accuracy by 14.2 percentage points.
• Score: 28.57781785937644
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Agentic LLMs exhibit three consequential tool-use failures: invalid arguments (validity), unnecessary calls (over-calling), and omitted calls when tools are needed (missing). We find that a small, failure-specific set of MLP neurons could distinguish such failures with linearly separable decision boundaries. Building on this observation, we introduce PRISMS (Probing Representations In Support of Monitoring and Steering), a closed-loop framework that shares a failure-specific neuron basis between sparse detection and activation steering. PRISMS selects contribution-critical MLP neurons and fits an L1-regularized detector on their activations. Across six models from the Qwen3, Llama, and Gemma families, over-calling and missing are detected at the pre-generation prompt boundary with ROC-AUC 0.90-1.00, while validity is detected from the generated tool-call span with ROC-AUC 0.86-0.90. These results are achieved with highly sparse readouts: only 1-2 MLP neurons for missing, 2-16 for over-calling, and approximately 128 for validity. These sparse detectors match or outperform dense residual-stream baselines using 23-627 times fewer features. The shared neuron basis also supports bidirectional control over tool-calling behavior, suppressing unnecessary calls and eliciting omitted ones. PRISMS therefore gates intervention on predicted failure risk to mitigate the collateral effects of unconditional steering. Across all six models, PRISMS reduces pooled over-calling rate by 80% (from 0.131 to 0.026) while increasing tool-required accuracy by 14.2 percentage points (from 0.689 to 0.831). PRISMS thus provides lightweight failure detection and selective intervention across model families.

Related papers

• Not All Relations Are Equal: Relation-Balanced and Calibrated Graph Learning for Provenance-Based Intrusion Detection [66.44527094471619]
  We present RECAL, an unsupervised framework using relation-balanced masked graph learning to better capture rare interaction patterns.<n>On three DARPA E3 datasets, RECAL achieves F1 scores of 99.99%, 99.93%, and 99.99%, outperforming the best baseline on each dataset by 0.88, 0.82, and 0.42 percentage points, respectively.
  arXiv  Detail & Related papers  (2026-09-15T00:31:01Z)
• It's the Problem, Not the Path: Budget and Difficulty Confounds in LLM Reasoning Trajectories [0.0]
  A restart-controlled truncation probe separates when a solution fits the continuation budget from when a prefix carries value that fresh computation cannot buy.<n>A pre-registered, difficulty-controlled test finds no detectable outcome information in early-window internal signals beyond a problem-difficulty baseline.
  arXiv  Detail & Related papers  (2026-09-03T06:47:55Z)
• Harnessing Disagreement: Detecting Correlated Agreement Blindness in Multi-Agent Triage [3.164789424470588]
  Disagreement-triggered escalation can create a structural blind spot in multi-agent arbitration.<n>We present ARAT, a directed-star system combining an inductive Random Forest (RF) agent, an analogical case-based k-nearest neighbour (k-NN) agent.
  arXiv  Detail & Related papers  (2026-07-22T08:35:00Z)
• Calibration Without Comprehension: Diagnosing the Limits of Fine-Tuning LLMs for Vulnerability Detection in Systems Software [1.0026496861838445]
  CWE-Trace is a framework for vulnerability detection built from 834 manually curated Linux kernel samples.<n>We evaluate eight vanilla LLMs and 15 LoRA fine-tuned variants across non-targeted detection, targeted detection, and CWE classification.
  arXiv  Detail & Related papers  (2026-06-18T17:19:26Z)
• Output Vector Editing for Memorization Mitigation in Large Language Models [68.30351930772788]
  Large language models memorize and reproduce sequences from their training data, creating privacy, copyright, and security risks.<n>Existing neuron-level mitigation methods equate editing with zeroing out neuron activations, but the activation only controls whether a neuron engages.<n>We propose output vector editing, a constrained-optimization edit that locates a small set of neurons responsible for a memorized continuation.
  arXiv  Detail & Related papers  (2026-06-17T07:29:18Z)
• Alignment Collapse Under KV Cache Quantization: Diagnosis and Mitigation [6.129872931808218]
  Key-value ( KV) cache quantization is widely used to reduce Large Language Model (LLM) inference memory.<n>In this study, we explore alignment preservation under KV cache quantization.<n>We find that low-bit quantization can silently destroy safety alignment.
  arXiv  Detail & Related papers  (2026-06-01T02:02:20Z)
• HunterAgent: Neuro-Symbolic Attack Trace Reconstruction under Anti-Forensics [4.5096964986324]
  Threat hunting requires reconstructing causal attack chains across heterogeneous logs.<n>We propose HunterAgent, a neurosymbolic framework that reframes trace reconstruction as cost-bounded graph search.<n>HunterAgent achieves 86.1% mean F1, outperforming the top agentic baseline by 26.7 F1 and KAIROS by 17.1 F1.
  arXiv  Detail & Related papers  (2026-05-28T02:38:12Z)
• Tracing Uncertainty in Language Model "Reasoning" [46.61902399979181]
  Language model (LM) "reasoning", commonly described as Chain-of-Thought or test-time scaling, often improves benchmark performance.<n>We study these dynamics by treating the "reasoning" traces, the intermediate token sequences generated by LMs, as evolving model states.
  arXiv  Detail & Related papers  (2026-05-08T14:16:37Z)
• HalluScan: A Systematic Benchmark for Detecting and Mitigating Hallucinations in Instruction-Following LLMs [0.0]
  Large Language Models (LLMs) have demonstrated remarkable capabilities across diverse natural language processing tasks.<n>Yet they remain susceptible to hallucinations -- generating content that is factually incorrect, unfaithful to provided context, or misaligned with user instructions.<n>We present HalluScan, a comprehensive benchmark framework that systematically evaluates hallucination detection and mitigation across 72 configurations.
  arXiv  Detail & Related papers  (2026-05-04T10:43:27Z)
• Hierarchy-Guided Topology Latent Flow for Molecular Graph Generation [44.50339042016925]
  We propose a planner-executor model that generates bond graphs with 3D coordinates using a latent multi-scale plan for global context.<n>HLTF achieves 98.8% atom stability and 92.9% valid-and-unique, improving PoseBusters validity to 94.0% (+0.9 over the strongest reported baseline)<n>On GEOM-DRUGS, HLTF attains 85.5%/85.0% validity/valid-unique-novel without post-processing and 92.2%/91.2% after standardized relaxation.
  arXiv  Detail & Related papers  (2026-03-28T03:48:13Z)
• BadCLIP++: Stealthy and Persistent Backdoors in Multimodal Contrastive Learning [73.46118996284888]
  Research on backdoor attacks against multimodal contrastive learning models faces two key challenges: stealthiness and persistence.<n>We propose BadCLIP++, a unified framework that tackles both challenges.<n>For stealthiness, we introduce a semantic-fusion QR micro-trigger that embeds imperceptible patterns near task-relevant regions.<n>For persistence, we stabilize trigger embeddings via radius shrinkage and centroid alignment.
  arXiv  Detail & Related papers  (2026-02-19T08:31:16Z)
• Information Fidelity in Tool-Using LLM Agents: A Martingale Analysis of the Model Context Protocol [69.11739400975445]
  We introduce the first theoretical framework for analyzing error accumulation in Model Context Protocol (MCP) agents.<n>We show that cumulative distortion exhibits linear growth and high-probability deviations bounded by $O(sqrtT)$.<n>Key findings include: semantic weighting reduces distortion by 80%, and periodic re-grounding approximately every 9 steps suffices for error control.
  arXiv  Detail & Related papers  (2026-02-10T21:08:53Z)
• On GRPO Collapse in Search-R1: The Lazy Likelihood-Displacement Death Spiral [59.14787085809595]
  We identify Lazy Likelihood Displacement (LLD) as the core mechanism driving this failure.<n>LLD emerges early and triggers a self-reinforcing LLD Death Spiral, where declining likelihood leads to low-confidence responses.<n>We propose a lightweight likelihood-preserving regularization LLDS for GRPO that activates only when a trajectory's likelihood decreases.
  arXiv  Detail & Related papers  (2025-12-03T19:41:15Z)

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.