To Call or Not to Call: Diagnosing Intrinsic Over-Calling Bias in LLM Agents
On the map: LLM agents and tool use → Tool learning and function calling
Abstract (arXiv)
LLM agents exhibit a consistent tendency to over-call, invoking tools even in situations where none is needed. On the When2Call benchmark, six models from three families show high call accuracy but much lower no-call accuracy, leaving overall accuracy in the 55%-70% range. We trace this to an Intrinsic Bias Hypothesis (IBH): the call/no-call decision mapping carries an activation-independent call offset, so the model favors call even at activation parity. Using Sparse Autoencoders (SAEs), we recover behavior-aligned feature bases for the call/no_call decision, reduce them to a signed activation margin, and estimate the offset directly. Across all six models, the model is decision-neutral only when no_call activation outweighs call activation, consistent with IBH. We then causally test IBH with Adaptive Margin-Calibrated Steering (AMCS), a closed-form counter-bias shift along SAE decoder directions. Cancelling the diagnosed offset mitigates over-calling and improves overall accuracy with a negligible drop in call accuracy. Our work recasts over-calling from an empirical phenomenon into a mechanistic object amenable to causal correction. Code is available at https://github.com/SKURA502/agent-sae/.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Tool learning and function calling, one of 16 directions in LLM agents and tool use. Builds on 18 of its 30 references that are in the map. First posted on arXiv 16 May 2026.
Builds on
18 of 30 references.
Closest ancestors · 4
Same direction, newest first
- 2025When2Call: When (not) to Call Tools26 Apr
- 2023ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs31 Jul
- 2023Gorilla: Large Language Model Connected with Massive APIs24 May
- 2023Toolformer: Language Models Can Teach Themselves to Use Tools9 Feb
Other references · 14
- 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
- 2020The Pile: An 800GB Dataset of Diverse Text for Language Modelingcited by 2,346
- 2025Gemma 3 Technical Reportcited by 1,686
- 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
- 2023AgentBench: Evaluating LLMs as Agentscited by 1,211
- 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,009
- 2024Scaling and evaluating sparse autoencoderscited by 530
- 2024AFlow: Automating Agentic Workflow Generationcited by 395
- 2024Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoderscited by 283
- 2025Open Problems in Mechanistic Interpretabilitycited by 173
- 2022MRKL Systems: A modular, neuro-symbolic architecture that combines large language models, external knowledge sources and discrete reasoningcited by 153
- 2026Ministral 3cited by 88
2 more
- 2025Route Sparse Autoencoder to Interpret Large Language Modelscited by 25
- 2025Interpretable Reward Model via Sparse Autoencodercited by 17
Built on by
Citing papers per year: 2026: 2.
2 papers.
Similar, not cited
Close in wording, no citation either way
- 2026Tunable Tool-Call Rates in LLM Agents via Representation Steering
- 2026A Few Neurons Reveal When LLMs Misuse Tools: Sparse Detection and Selective Steering for Reliable Tool Use
- 2026Evaluate the Stack, Not the Layer: Do Deterministic and LLM Gates for Agent Actions Fail Independently?
- 2026Tool-Call Dependency Structure is Linearly Decodable in LLM Agent Residual Streams
- 2026Proxy Confidence: Auditing Black-Box LLM Agents with a Surrogate's Log-Probabilities
- 2026A rigor-matched audit of periodic-step layer skipping for efficient llm inference: conflayers versus swift, with a supplemental analysis of trained routing alternatives
- 2026From Confident Closing to Silent Failure: Characterizing False Success in LLM Agents
- 2026Fast Models, Slow Evidence: A Paired and Self-Audited Evaluation of System-1 Decision Models for LLM Agent Harnesses