mlsanityData to 7 Oct 2026

Nudging Hidden States: Training-Free Model Steering for Chain-of-Thought Reasoning in Large Audio-Language Models

arXiv 2603.14636 ↗First posted 15 Mar 2026cs.SD cs.AI cs.CLcited by 1

On the map: LLM reasoning → Audio-language model reasoning

Abstract (arXiv)

Chain-of-thought (CoT) prompting has been extended to large audio-language models (LALMs) to elicit reasoning, yet enhancing its effectiveness without training remains challenging. We study inference-time model steering as a training-free approach to improve LALM reasoning. We introduce three strategies using diverse information sources and evaluate them across four LALMs and four benchmarks. Results show general accuracy gains up to 4.4% over CoT prompting. Notably, we identify a cross-modal transfer where steering vectors derived from few text samples effectively guide speech-based reasoning, demonstrating high data efficiency. We also examine hyperparameter sensitivity to understand the robustness of these approaches. Our findings position model steering as a practical direction for strengthening LALM reasoning.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Audio-language model reasoning, one of 17 directions in LLM reasoning. Builds on 49 of its 52 references that are in the map. First posted on arXiv 15 Mar 2026.

Builds on

49 of 52 references.

Closest ancestors · 9

Same direction, newest first

  1. 2025When Silence Matters: The Impact of Irrelevant Audio on Text Reasoning in Large Audio-Language Models1 Oct
  2. 2025Audio-Thinker: Guiding Audio Language Model When and How to Think via Reinforcement Learning11 Aug
  3. 2025SpeechR: A Benchmark for Speech Reasoning in Large Audio-Language Models4 Aug
  4. 2025Audio Flamingo 3: Advancing Audio Intelligence with Fully Open Large Audio Language Models10 Jul
  5. 2025SAKURA: On the Multi-hop Reasoning of Large Audio-Language Models Based on Speech and Audio Information19 May
  6. 2025Qwen2.5-Omni Technical Report26 Mar
  7. 2025Audio-Reasoner: Improving Reasoning Capability in Large Audio Language Models4 Mar
  8. 2025Audio-CoT: Exploring Chain-of-Thought Reasoning in Large Audio Language Model13 Jan
  9. 2024MMAU: A Massive Multi-Task Audio Understanding and Reasoning Benchmark24 Oct

Other references · 40

  1. 2024The Llama 3 Herd of Modelscited by 14,192
  2. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  3. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  4. 2025Qwen3 Technical Reportcited by 8,195
  5. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  6. 2024GPT-4o System Cardcited by 4,467
  7. 2022Large Language Models are Zero-Shot Reasonerscited by 4,323
  8. 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,009
  9. 2024Refusal in Language Models Is Mediated by a Single Directioncited by 960
  10. 2023Steering Language Models With Activation Engineeringcited by 824
  11. 2023The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasetscited by 673
  12. 2025Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAscited by 491
28 more
  1. 2022Extracting Latent Steering Vectors from Pretrained Language Modelscited by 217
  2. 2024Improving Instruction-Following in Language Models through Activation Steeringcited by 149
  3. 2025On The Landscape of Spoken Language Models: A Comprehensive Surveycited by 116
  4. 2025IndexTTS2: A Breakthrough in Emotionally Expressive and Duration-Controlled Auto-Regressive Zero-Shot Text-to-Speechcited by 85
  5. 2024Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Taskscited by 70
  6. 2025Understanding Reasoning in Thinking Language Models via Steering Vectorscited by 69
  7. 2025MMAU-Pro: A Challenging and Comprehensive Benchmark for Holistic Evaluation of Audio General Intelligencecited by 59
  8. 2025DeSTA2.5-Audio: Toward General-Purpose Large Audio Language Model with Self-Generated Cross-Modal Alignmentcited by 47
  9. 2025Voxtralcited by 41
  10. 2025Understanding (Un)Reliability of Steering Vectors in Language Modelscited by 39
  11. 2025Towards Holistic Evaluation of Large Audio-Language Models: A Comprehensive Surveycited by 38
  12. 2025Mitigating Overthinking in Large Reasoning Models via Manifold Steeringcited by 32
  13. 2024Speech-Copilot: Leveraging Large Language Models for Speech Processing via Task Decomposition, Modularization, and Program Generationcited by 27
  14. 2025Speech-IFEval: Evaluating Instruction-Following and Quantifying Catastrophic Forgetting in Speech-Aware Language Modelscited by 27
  15. 2024Building a Taiwanese Mandarin Spoken Language Model: A First Attemptcited by 22
  16. 2025Mitigating Content Effects on Reasoning in Language Models through Fine-Grained Activation Steeringcited by 18
  17. 2025AudioLens: A Closer Look at Auditory Attribute Perception of Large Audio-Language Modelscited by 18
  18. 2025VoxEval: Benchmarking the Knowledge Understanding Capabilities of End-to-End Spoken Language Modelscited by 16
  19. 2025A Preliminary Exploration with GPT-4o Voice Modecited by 16
  20. 2025Steering LLM Reasoning Through Bias-Only Adaptationcited by 16
  21. 2025Analyzing Mitigation Strategies for Catastrophic Forgetting in End-to-End Training of Spoken Language Modelscited by 11
  22. 2025Improving LLM Reasoning through Interpretable Role-Playing Steeringcited by 11
  23. 2025BILLY: Steering Large Language Models via Merging Persona Vectors for Creative Generationcited by 10
  24. 2025Silence is Golden: Mitigating Hallucinations in Large Audio-Language Models via Layer-Weighted Vector Steeringcited by 6
  25. 2025SARSteer: Safeguarding Large Audio-Language Models via Safe-Ablated Refusal Steeringcited by 6
  26. 2025SAKE: Towards Editing Auditory Attribute Knowledge of Large Audio-Language Modelscited by 5
  27. 2025Towards General Auditory Intelligence: Large Multimodal Models for Machine Listening and Speakingcited by 4
  28. 2025Self-Improving Model Steeringcited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Encoder-Side Neuron Identification and Amplification for Acoustic Perception in Large Audio-Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Thinking with Sound: Audio Chain-of-Thought Enables Multimodal Reasoning in Large Audio-Language Models
  2. 2025Think Smart, Not Hard: Difficulty Adaptive Reasoning for Large Audio Language Models
  3. 2026X3-OPD: Distilling Reasoning into Large Audio-Language Models via On-Policy Alignment
  4. 2024Chain-of-Thought Augmentation with Logit Contrast for Enhanced Reasoning in Language Models
  5. 2025Can Speech LLMs Think while Listening?
  6. 2025Prototype-Based Dynamic Steering for Large Language Models
  7. 2025SHANKS: Simultaneous Hearing and Thinking for Spoken Language Models
  8. 2026Listen Then Reason: Perception-Grounded Test-Time Reinforcement Learning for Large Audio-Language Models