# The First to Know: How Token Distributions Reveal Hidden Knowledge in Large Vision-Language Models?

arXiv: https://arxiv.org/abs/2403.09037 · HTML: https://mlsanity.com/paper/2403.09037/

## Key facts

- Cited by 26 papers in the map (data through 14 Oct 2026).
- Placed in Hallucination mitigation in LVLMs, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 30 of its 45 references that are in the map.
- First posted on arXiv 14 Mar 2024.

## Placed in

- Multimodal LLMs / vision-language models / [Hallucination mitigation in LVLMs](https://mlsanity.com/mllm/hallucination-mitigation-in-lvlms/)
- LLM safety and security / [Machine unlearning](https://mlsanity.com/safety/machine-unlearning/)

## Builds on (most cited first, 30 of 30)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis](https://mlsanity.com/paper/2307.01952/) (2023)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models](https://mlsanity.com/paper/2304.10592/) (2023)
- [MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts](https://mlsanity.com/paper/2310.02255/) (2023)
- [Representation Engineering: A Top-Down Approach to AI Transparency](https://mlsanity.com/paper/2310.01405/) (2023)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [VizWiz Grand Challenge: Answering Visual Questions from Blind People](https://mlsanity.com/paper/1802.08218/) (2018)
- [Discovering Latent Knowledge in Language Models Without Supervision](https://mlsanity.com/paper/2212.03827/) (2022)
- [The Dawn of LMMs: Preliminary Explorations with GPT-4V(ision)](https://mlsanity.com/paper/2309.17421/) (2023)
- [LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Model](https://mlsanity.com/paper/2304.15010/) (2023)
- [mPLUG-Owl2: Revolutionizing Multi-modal Large Language Model with Modality Collaboration](https://mlsanity.com/paper/2311.04257/) (2023)
- [Towards Measuring the Representation of Subjective Global Opinions in Language Models](https://mlsanity.com/paper/2306.16388/) (2023)
- [Language Models Represent Space and Time](https://mlsanity.com/paper/2310.02207/) (2023)
- [MultiModal-GPT: A Vision and Language Model for Dialogue with Humans](https://mlsanity.com/paper/2305.04790/) (2023)
- [Do LLMs exhibit human-like response biases? A case study in survey design](https://mlsanity.com/paper/2311.04076/) (2023)
- [MLLM-Protector: Ensuring MLLM's Safety without Hurting Performance](https://mlsanity.com/paper/2401.02906/) (2024)
- [DRESS: Instructing Large Vision-Language Models to Align and Interact with Humans via Natural Language Feedback](https://mlsanity.com/paper/2311.10081/) (2023)
- [The Curious Case of Hallucinatory (Un)answerability: Finding Truths in the Hidden States of Over-Confident Large Language Models](https://mlsanity.com/paper/2310.11877/) (2023)
- [How Easy is It to Fool Your Multimodal LLMs? An Empirical Analysis on Deceptive Prompts](https://mlsanity.com/paper/2402.13220/) (2024)
- [Open-ended VQA benchmarking of Vision-Language models by exploiting Classification datasets and their semantic hierarchy](https://mlsanity.com/paper/2402.07270/) (2024)
- [MM-SAP: A Comprehensive Benchmark for Assessing Self-Awareness of Multimodal Large Language Models in Perception](https://mlsanity.com/paper/2401.07529/) (2024)
- [Towards Tracing Trustworthiness Dynamics: Revisiting Pre-training Period of Large Language Models](https://mlsanity.com/paper/2402.19465/) (2024)

## Built on by (most cited first, 26 of 26)

- [Cross-modal Information Flow in Multimodal Large Language Models](https://mlsanity.com/paper/2411.18620/) (2024)
- [Explainable and Interpretable Multimodal Large Language Models: A Comprehensive Survey](https://mlsanity.com/paper/2412.02104/) (2024)
- [Jailbreaking Multimodal Large Language Models via Shuffle Inconsistency](https://mlsanity.com/paper/2501.04931/) (2025)
- [Visual Large Language Models for Generalized and Specialized Applications](https://mlsanity.com/paper/2501.02765/) (2025)
- [Unbridled Icarus: A Survey of the Potential Perils of Image Inputs in Multimodal Large Language Model Security](https://mlsanity.com/paper/2404.05264/) (2024)
- [Vision Function Layer in Multimodal LLMs](https://mlsanity.com/paper/2509.24791/) (2025)
- [LLM Microscope: What Model Internals Reveal About Answer Correctness and Context Utilization](https://mlsanity.com/paper/2510.04013/) (2025)
- [Training-Free Personalization via Retrieval and Reasoning on Fingerprints](https://mlsanity.com/paper/2503.18623/) (2025)
- [Attention Hijackers: Detect and Disentangle Attention Hijacking in LVLMs for Hallucination Mitigation](https://mlsanity.com/paper/2503.08216/) (2025)
- [Collaborative Instance Object Navigation: Leveraging Uncertainty-Awareness to Minimize Human-Agent Dialogues](https://mlsanity.com/paper/2412.01250/) (2024)
- [Legend: Leveraging Representation Engineering to Annotate Safety Margin for Preference Datasets](https://mlsanity.com/paper/2406.08124/) (2024)
- [When Sinks Help or Hurt: Unified Framework for Attention Sink in Large Vision-Language Models](https://mlsanity.com/paper/2604.03316/) (2026)
- [MTRE: Multi-Token Reliability Estimation for Hallucination Detection in VLMs](https://mlsanity.com/paper/2505.11741/) (2025)
- [HARMONY: Hidden Activation Representations and Model Output-Aware Uncertainty Estimation for Vision-Language Models](https://mlsanity.com/paper/2510.22171/) (2025)
- [Contextual Image Attack: How Visual Context Exposes Multimodal Safety Vulnerabilities](https://mlsanity.com/paper/2512.02973/) (2025)
- [Where Does Vision Meet Language? Understanding and Refining Visual Fusion in MLLMs via Contrastive Attention](https://mlsanity.com/paper/2601.08151/) (2026)
- [Hallucination as Commitment Failure: Larger LLMs Misfire Despite Knowing the Answer](https://mlsanity.com/paper/2605.22007/) (2026)
- [Topological Signatures of Adversaries in Multimodal Alignments](https://mlsanity.com/paper/2501.18006/) (2025)
- [TEMPO: Learning Temporal Context for Dynamic Robot Manipulation](https://mlsanity.com/paper/2609.16864/) (2026)
- [Safety-Potential Pruning for Enhancing Safety Prompts Against VLM Jailbreaking Without Retraining](https://mlsanity.com/paper/2603.14219/) (2026)
- [UHP Detection: LVLMs have their Unique Hallucination Pattern in the Consistency Space](https://mlsanity.com/paper/2608.03817/) (2026)
- [Uncertainty-Aware Decision Making in Multimodal Large Language Models](https://mlsanity.com/paper/2608.17084/) (2026)
- [From Knowing to Abstaining: Bridging the Representation-Action Gap in Vision-Language Models](https://mlsanity.com/paper/2609.32653/) (2026)
- [The Commit-Abstain Circuit: Why Language Models Hallucinate Instead of Abstaining](https://mlsanity.com/paper/2609.32964/) (2026)
- [COMPASS: Finding Where Reasoning Lives in Language Models](https://mlsanity.com/paper/2610.07469/) (2026)
- [RepProbe: Can Video Language Models Count Repeated Actions? A Diagnostic Benchmark](https://mlsanity.com/paper/2610.18844/) (2026)

## Abstract (arXiv)

Large vision-language models (LVLMs), designed to interpret and respond to human instructions, occasionally generate hallucinated or harmful content due to inappropriate instructions. This study uses linear probing to shed light on the hidden knowledge at the output layers of LVLMs. We demonstrate that the logit distributions of the first tokens contain sufficient information to determine whether to respond to the instructions, including recognizing unanswerable visual questions, defending against jailbreaking attacks, and identifying deceptive questions. Such hidden knowledge is gradually lost in logits of subsequent tokens during response generation. Then, we illustrate a simple decoding strategy at the generation of the first token, effectively improving the generated content. In experiments, we find a few interesting insights: First, the CLIP model already contains a strong signal for solving these tasks, which indicates potential bias in the existing datasets. Second, we observe performance improvement by utilizing the first logit distributions on three additional tasks, including indicating uncertainty in math solving, mitigating hallucination, and image classification. Last, with the same training data, simply finetuning LVLMs improves models' performance but is still inferior to linear probing on these tasks.