mlsanityData to 7 Oct 2026

MINER: Mining the Underlying Pattern of Modality-Specific Neurons in Multimodal Large Language Models

arXiv 2410.04819 ↗First posted 7 Oct 2024cs.CLcited by 7

On the map: Multimodal LLMs / vision-language models → MLLM training and visual instruction tuning

Abstract (arXiv)

In recent years, multimodal large language models (MLLMs) have significantly advanced, integrating more modalities into diverse applications. However, the lack of explainability remains a major barrier to their use in scenarios requiring decision transparency. Current neuron-level explanation paradigms mainly focus on knowledge localization or language- and domain-specific analyses, leaving the exploration of multimodality largely unaddressed. To tackle these challenges, we propose MINER, a transferable framework for mining modality-specific neurons (MSNs) in MLLMs, which comprises four stages: (1) modality separation, (2) importance score calculation, (3) importance score aggregation, (4) modality-specific neuron selection. Extensive experiments across six benchmarks and two representative MLLMs show that (I) deactivating ONLY 2% of MSNs significantly reduces MLLMs performance (0.56 to 0.24 for Qwen2-VL, 0.69 to 0.31 for Qwen2-Audio), (II) different modalities mainly converge in the lower layers, (III) MSNs influence how key information from various modalities converges to the last token, (IV) two intriguing phenomena worth further investigation, i.e., semantic probing and semantic telomeres. The source code is available at this URL.

Key facts. Cited by 7 papers in the map (data through 7 Oct 2026). Placed in MLLM training and visual instruction tuning, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 50 of its 62 references that are in the map. First posted on arXiv 7 Oct 2024.

Builds on

50 of 62 references.

Closest ancestors · 7

Same direction, newest first

  1. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution18 Sep
  2. 2024MMNeuron: Discovering Neuron-Level Domain-Specific Interpretation in Multimodal Large Language Model17 Jun
  3. 2024A Concept-Based Explainability Framework for Large Multimodal Models12 Jun
  4. 2024AnyGPT: Unified Multimodal LLM with Discrete Sequence Modeling19 Feb
  5. 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks21 Dec
  6. 2023Finding and Editing Multi-Modal Neurons in Pre-Trained Transformers13 Nov
  7. 2023MMBench: Is Your Multi-modal Model an All-around Player?12 Jul

Other references · 43

  1. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  2. 2023GPT-4 Technical Reportcited by 18,132
  3. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  4. 2024The Llama 3 Herd of Modelscited by 14,192
  5. 2023Visual Instruction Tuningcited by 7,792
  6. 2020Measuring Massive Multitask Language Understandingcited by 7,768
  7. 2023BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Modelscited by 6,069
  8. 2023Qwen Technical Reportcited by 3,273
  9. 2022Locating and Editing Factual Associations in GPTcited by 2,609
  10. 2023Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyondcited by 2,179
  11. 2024Qwen2 Technical Reportcited by 2,076
  12. 2019Language Models as Knowledge Bases?cited by 1,922
31 more
  1. 2019Towards VQA Models That Can Readcited by 1,696
  2. 2020Transformer Feed-Forward Layers Are Key-Value Memoriescited by 1,307
  3. 2022Mass-Editing Memory in a Transformercited by 976
  4. 2024DeepSeek-VL: Towards Real-World Vision-Language Understandingcited by 718
  5. 2017Network Dissection: Quantifying Interpretability of Deep Visual Representationscited by 703
  6. 2023Qwen-Audio: Advancing Universal Audio Understanding via Unified Large-Scale Audio-Language Modelscited by 591
  7. 2021Knowledge Neurons in Pretrained Transformerscited by 590
  8. 2024Qwen2-Audio Technical Reportcited by 543
  9. 2024Scaling and evaluating sparse autoencoderscited by 530
  10. 2023SALMONN: Towards Generic Hearing Abilities for Large Language Modelscited by 492
  11. 2019A Survey on Explainable Artificial Intelligence (XAI): Towards Medical XAIcited by 403
  12. 2023Explainability for Large Language Models: A Surveycited by 331
  13. 2024Do Llamas Work in English? On the Latent Language of Multilingual Transformerscited by 275
  14. 2023Does Localization Inform Editing? Surprising Differences in Causality-Based Localization vs. Knowledge Editing in Language Modelscited by 237
  15. 2023Pengi: An Audio Language Model for Audio Taskscited by 217
  16. 2023Any-to-Any Generation via Composable Diffusioncited by 203
  17. 2022CLIP-Dissect: Automatic Description of Neuron Representations in Deep Vision Networkscited by 144
  18. 2023On decoder-only architecture for speech-to-text and large language model integrationcited by 144
  19. 2024Language-Specific Neurons: The Key to Multilingual Capabilities in Large Language Modelscited by 124
  20. 2019Augmenting Self-attention with Persistent Memorycited by 92
  21. 2024On the Multilingual Ability of Decoder-based Pre-trained Language Models: Finding and Controlling Language-Specific Neuronscited by 83
  22. 2023UrbanCLIP: Learning Text-enhanced Urban Region Profiling with Contrastive Language-Image Pretraining from the Webcited by 77
  23. 2022Vocalsound: A Dataset for Improving Human Vocal Sounds Recognitioncited by 62
  24. 2024SpeechVerse: A Large-scale Generalizable Audio Language Modelcited by 62
  25. 2023Journey to the Center of the Knowledge Neurons: Discoveries of Language-Independent Knowledge Neurons and Degenerate Knowledge Neuronscited by 56
  26. 2023Multimodal Neurons in Pretrained Text-Only Transformerscited by 49
  27. 2024A Comprehensive Survey of Large Language Models and Multimodal Large Language Models in Medicinecited by 45
  28. 2023Multimodal Explainable Artificial Intelligence: A Comprehensive Review of Methodological Advances and Future Research Directionscited by 25
  29. 2023Localized Symbolic Knowledge Distillation for Visual Commonsense Modelscited by 12
  30. 2024MIND: Multimodal Shopping Intention Distillation from Large Vision-language Models for E-commerce Purchase Understandingcited by 10
  31. 2024FastMem: Fast Memorization of Prompt Improves Context Awareness of Large Language Modelscited by 7

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 1; 2025: 1; 2026: 5.

7 papers.

  1. 2024Explainable and Interpretable Multimodal Large Language Models: A Comprehensive Surveycited by 52
  2. 2025Analyzing Finetuning Representation Shift for Multimodal LLMs Steeringcited by 15
  3. 2026Instruction Anchor: Dissecting the Mechanistic Dynamics of Modality Arbitrationcited by 6
  4. 2026Interpretable Debiasing of Vision-Language Models for Social Fairnesscited by 6
  5. 2026From Heads to Neurons: Causal Attribution and Steering in Multi-Task Vision-Language Modelscited by 0
  6. 2026Emotion Across Speech and Faces: Shared Affective Mechanisms in Multimodal Foundation Modelscited by 0
  7. 2026Who Is Left of Whom? Tracing Spatial Evidence and Role Binding in Relative-Position Reasoningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Modality-Aware Neuron Pruning for Unlearning in Multimodal Large Language Models
  2. 2026Correct When Paired, Wrong When Split: Decoupling and Editing Modality-Specific Neurons in MLLMs
  3. 2025Cross-Modal Unlearning via Influential Neuron Path Editing in Multimodal Large Language Models
  4. 2025Identifying Multi-modal Knowledge Neurons in Pretrained Transformers via Two-stage Filtering
  5. 2024Model Composition for Multimodal Large Language Models
  6. 2026Mixture of Probes: Learning from Privileged Modalities in Multimodal LLMs Through Probing
  7. 2026MoST: Mixing Speech and Text with Modality-Aware Mixture of Experts
  8. 2024A Comprehensive Review of Multimodal Large Language Models: Performance and Challenges Across Different Tasks