When Does Correction Become Repair? Mechanistic Auditing of Internal Interventions in Tool-Using LLMs
cs.CL, cs.SE
Submitted: 2026-09-28
Updated: 2026-09-28
Code: https://github.com/ruizheliUOA/mechanistic-tool-use-llm
Terminology
Sources
- Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone
- ToolRM: Outcome Reward Models for Tool-Calling Large Language Models
- The Llama 3 Herd of Models
- MetaTool Benchmark for Large Language Models: Deciding Whether to Use Tools and Which to Use
- From Text to Voice: A Reproducible and Verifiable Framework for Evaluating Tool Calling LLM Agents
- AgentAbstain: Do LLM Agents Know When Not to Act?
- Qwen2.5 Technical Report
- To Call or Not to Call: Diagnosing Intrinsic Over-Calling Bias in LLM Agents
- ToolAlpaca: Generalized Tool Learning for Language Models with 3000 Simulated Cases
- Beyond the Black Box: Interpretability of Agentic AI Tool Use
- Gemma 2: Improving Open Language Models at a Practical Size
- Position: Agent Should Invoke External Tools ONLY When Epistemically Necessary
- ASA: Backbone-Training-Free Representation Engineering for Tool-Calling Agents
- To Call or Not to Call: A Framework to Assess and Optimize LLM Tool Calling
- Qwen3 Technical Report
- Calibration is the Bottleneck: An Action-Class Diagnostic of Multi-Turn Tool-Calling
- Exploring Agentic Tool-Calling Decisions via Uncertainty-Aligned Reinforcement Learning
- Representation Engineering: A Top-Down Approach to AI Transparency
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering