# Debiasing Multimodal Large Language Models via Penalization of Language Priors

arXiv: https://arxiv.org/abs/2403.05262 · HTML: https://mlsanity.com/paper/2403.05262/

## Key facts

- Cited by 37 papers in the map (data through 14 Oct 2026).
- Placed in MLLM training and visual instruction tuning, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 23 of its 51 references that are in the map.
- First posted on arXiv 8 Mar 2024.

## Placed in

- Multimodal LLMs / vision-language models / [MLLM training and visual instruction tuning](https://mlsanity.com/mllm/mllm-training-and-visual-instruction-tuning/)

## Builds on (most cited first, 23 of 23)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [Qwen Technical Report](https://mlsanity.com/paper/2309.16609/) (2023)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [Emergent Abilities of Large Language Models](https://mlsanity.com/paper/2206.07682/) (2022)
- [Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond](https://mlsanity.com/paper/2308.12966/) (2023)
- [MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI](https://mlsanity.com/paper/2311.16502/) (2023)
- [Evaluating Object Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2305.10355/) (2023)
- [MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models](https://mlsanity.com/paper/2306.13394/) (2023)
- [Aligning Large Multimodal Models with Factually Augmented RLHF](https://mlsanity.com/paper/2309.14525/) (2023)
- [Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding](https://mlsanity.com/paper/2311.16922/) (2023)
- [Object Hallucination in Image Captioning](https://mlsanity.com/paper/1809.02156/) (2018)
- [Contrastive Decoding: Open-ended Text Generation as Optimization](https://mlsanity.com/paper/2210.15097/) (2022)
- [Mitigating Hallucination in Large Multi-Modal Models via Robust Instruction Tuning](https://mlsanity.com/paper/2306.14565/) (2023)
- [Trusting Your Evidence: Hallucinate Less with Context-aware Decoding](https://mlsanity.com/paper/2305.14739/) (2023)
- [Analyzing and Mitigating Object Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2310.00754/) (2023)
- [A Comprehensive Survey of Hallucination Mitigation Techniques in Large Language Models](https://mlsanity.com/paper/2401.01313/) (2024)
- [Detecting and Preventing Hallucinations in Large Vision Language Models](https://mlsanity.com/paper/2308.06394/) (2023)
- [Online Deep Learning: Learning Deep Neural Networks on the Fly](https://mlsanity.com/paper/1711.03705/) (2017)
- [Context-faithful Prompting for Large Language Models](https://mlsanity.com/paper/2303.11315/) (2023)
- [Towards End-to-End Embodied Decision Making via Multi-modal Large Language Model: Explorations with GPT4-Vision and Beyond](https://mlsanity.com/paper/2310.02071/) (2023)
- [Speculative Contrastive Decoding](https://mlsanity.com/paper/2311.08981/) (2023)

## Built on by (most cited first, 30 of 37)

- [Thyme: Think Beyond Images](https://mlsanity.com/paper/2508.11630/) (2025)
- [MLLM can see? Dynamic Correction Decoding for Hallucination Mitigation](https://mlsanity.com/paper/2410.11779/) (2024)
- [MM-RLHF: The Next Step Forward in Multimodal LLM Alignment](https://mlsanity.com/paper/2502.10391/) (2025)
- [R1-Reward: Training Multimodal Reward Model Through Stable Reinforcement Learning](https://mlsanity.com/paper/2505.02835/) (2025)
- [MLLMs are Deeply Affected by Modality Bias](https://mlsanity.com/paper/2505.18657/) (2025)
- [AutoBench-V: Can Large Vision-Language Models Benchmark Themselves?](https://mlsanity.com/paper/2410.21259/) (2024)
- [VLM-Guard: Safeguarding Vision-Language Models via Fulfilling Safety Alignment Gap](https://mlsanity.com/paper/2502.10486/) (2025)
- [Aligning Multimodal LLM with Human Preference: A Survey](https://mlsanity.com/paper/2503.14504/) (2025)
- [VTC-Bench: Evaluating Agentic Multimodal Models via Compositional Visual Tool Chaining](https://mlsanity.com/paper/2603.15030/) (2026)
- [DatBench: Discriminative, Faithful, and Efficient VLM Evaluations](https://mlsanity.com/paper/2601.02316/) (2026)
- [OmniDPO: A Preference Optimization Framework to Address Omni-Modal Hallucination](https://mlsanity.com/paper/2509.00723/) (2025)
- [RITUAL: Random Image Transformations as a Universal Anti-hallucination Lever in Large Vision Language Models](https://mlsanity.com/paper/2405.17821/) (2024)
- [BrowseComp-V3: A Visual, Vertical, and Verifiable Benchmark for Multimodal Browsing Agents](https://mlsanity.com/paper/2602.12876/) (2026)
- [LatentOmni: Rethinking Omni-Modal Understanding via Unified Audio-Visual Latent Reasoning](https://mlsanity.com/paper/2605.22012/) (2026)
- [Self-Reflective Planning with Knowledge Graphs: Enhancing LLM Reasoning Reliability for Question Answering](https://mlsanity.com/paper/2505.19410/) (2025)
- [Investigating VLM Hallucination from a Cognitive Psychology Perspective: A First Step Toward Interpretation with Intriguing Observations](https://mlsanity.com/paper/2507.03123/) (2025)
- [Are Audio-Language Models Listening? Audio-Specialist Heads for Adaptive Audio Steering](https://mlsanity.com/paper/2603.06854/) (2026)
- [Artifact-Bench: Evaluating MLLMs on Detecting and Assessing the Artifacts of AI-Generated Videos](https://mlsanity.com/paper/2605.18984/) (2026)
- [Challenges in Understanding Modality Conflict in Vision-Language Models](https://mlsanity.com/paper/2509.02805/) (2025)
- [Focusing Where Vision Matters: Selective Training for Large Vision Language Models via Visual Information Gain](https://mlsanity.com/paper/2602.17186/) (2026)
- [Beyond the Last Layer: Multi-Layer Representation Fusion for Visual Tokenization](https://mlsanity.com/paper/2605.10780/) (2026)
- [OmniPack: Unified Token Compression for Efficient Omni-modal Large Language Models](https://mlsanity.com/paper/2608.03812/) (2026)
- [On the Risk of Misleading Reports: Diagnosing Textual Biases in Multimodal Clinical AI](https://mlsanity.com/paper/2508.00171/) (2025)
- [Unleashing the Intrinsic Visual Representation Capability of Multimodal Large Language Models](https://mlsanity.com/paper/2512.06281/) (2025)
- [SAVAA: Mitigating Hallucinations in LVLMs via Step-wise Adaptive Visual Attention Amplification](https://mlsanity.com/paper/2602.13600/) (2026)
- [SAGE: Sink-Aware Grounded Decoding for Multimodal Hallucination Mitigation](https://mlsanity.com/paper/2603.27898/) (2026)
- [Debias in Text, Believe Your Eyes: Text-Anchored Cross-Modal Transfer for Visual Counter-Commonsense Reasoning](https://mlsanity.com/paper/2608.06938/) (2026)
- [Understanding Temporal Logic Consistency in Video-Language Models through Cross-Modal Attention Discriminability](https://mlsanity.com/paper/2510.08138/) (2025)
- [Hybrid Attribution Priors for Explainable and Robust Model Training](https://mlsanity.com/paper/2512.14719/) (2025)
- [Self-Correction Inside the Model: Leveraging Layer Attention to Mitigate Hallucinations in Large Vision Language Models](https://mlsanity.com/paper/2603.00437/) (2026)

## Abstract (arXiv)

In the realms of computer vision and natural language processing, Multimodal Large Language Models (MLLMs) have become indispensable tools, proficient in generating textual responses based on visual inputs. Despite their advancements, our investigation reveals a noteworthy bias: the generated content is often driven more by the inherent priors of the underlying Large Language Models (LLMs) than by the input image. Empirical experiments underscore the persistence of this bias, as MLLMs often provide confident answers even in the absence of relevant images or given incongruent visual inputs. To rectify these biases and redirect the model's focus toward visual information, we propose two simple, training-free strategies. First, for tasks such as classification or multi-choice question answering, we introduce a "Post-Hoc Debias" method using an affine calibration step to adjust the output distribution. This approach ensures uniform answer scores when the image is absent, acting as an effective regularization technique to alleviate the influence of LLM priors. For more intricate open-ended generation tasks, we extend this method to "Visual Debias Decoding", which mitigates bias by contrasting token log-probabilities conditioned on a correct image versus a meaningless one. Additionally, our investigation sheds light on the instability of MLLMs across various decoding configurations. Through systematic exploration of different settings, we achieve significant performance improvements--surpassing previously reported results--and raise concerns about the fairness of current evaluation practices. Comprehensive experiments substantiate the effectiveness of our proposed strategies in mitigating biases. These strategies not only prove beneficial in minimizing hallucinations but also contribute to the generation of more helpful and precise illustrations.