mlsanityData to 7 Oct 2026

Stop Reasoning! When Multimodal LLM with Chain-of-Thought Reasoning Meets Adversarial Image

arXiv 2402.14899 ↗First posted 22 Feb 2024cs.CV cs.AI cs.CRcited by 26

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

Multimodal LLMs (MLLMs) with a great ability of text and image understanding have received great attention. To achieve better reasoning with MLLMs, Chain-of-Thought (CoT) reasoning has been widely explored, which further promotes MLLMs' explainability by giving intermediate reasoning steps. Despite the strong power demonstrated by MLLMs in multimodal reasoning, recent studies show that MLLMs still suffer from adversarial images. This raises the following open questions: Does CoT also enhance the adversarial robustness of MLLMs? What do the intermediate reasoning steps of CoT entail under adversarial attacks? To answer these questions, we first generalize existing attacks to CoT-based inferences by attacking the two main components, i.e., rationale and answer. We find that CoT indeed improves MLLMs' adversarial robustness against the existing attack methods by leveraging the multi-step reasoning process, but not substantially. Based on our findings, we further propose a novel attack method, termed as stop-reasoning attack, that attacks the model while bypassing the CoT reasoning process. Experiments on three MLLMs and two visual reasoning datasets verify the effectiveness of our proposed method. We show that stop-reasoning attack can result in misled predictions and outperform baseline attacks by a significant margin.

Key facts. Cited by 26 papers in the map (data through 7 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 32 of its 36 references that are in the map. First posted on arXiv 22 Feb 2024.

Builds on

32 of 36 references.

Closest ancestors · 4

Same direction, newest first

  1. 2023Multi-modal Latent Space Learning for Chain-of-Thought Reasoning in Language Models14 Dec
  2. 2023How Many Unicorns Are in This Image? A Safety Evaluation Benchmark for Vision LLMs27 Nov
  3. 2023Multimodal Chain-of-Thought Reasoning in Language Models2 Feb
  4. 2022Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering20 Sep

Other references · 28

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  2. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  3. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  4. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  5. 2013Intriguing properties of neural networkscited by 5,114
  6. 2023Improved Baselines with Visual Instruction Tuningcited by 4,362
  7. 2022Large Language Models are Zero-Shot Reasonerscited by 4,323
  8. 2016Towards Evaluating the Robustness of Neural Networkscited by 2,976
  9. 2023Universal and Transferable Adversarial Attacks on Aligned Language Modelscited by 2,914
  10. 2023MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Modelscited by 2,689
  11. 2022Large Language Models Encode Clinical Knowledgecited by 1,807
  12. 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,259
16 more
  1. 2022A-OKVQA: A Benchmark for Visual Question Answering using World Knowledgecited by 702
  2. 2022Automatic Chain of Thought Prompting in Large Language Modelscited by 684
  3. 2023OpenFlamingo: An Open-Source Framework for Training Large Autoregressive Vision-Language Modelscited by 560
  4. 2019On Evaluating Adversarial Robustnesscited by 442
  5. 2020Large-Scale Adversarial Training for Vision-and-Language Representation Learningcited by 327
  6. 2023On Evaluating Adversarial Robustness of Large Vision-Language Modelscited by 281
  7. 2023MM-SafetyBench: A Benchmark for Safety Evaluation of Multimodal Large Language Modelscited by 255
  8. 2023Certifying LLM Safety against Adversarial Promptingcited by 248
  9. 2023Image Hijacks: Adversarial Images can Control Generative Models at Runtimecited by 165
  10. 2021Interpretable Deep Learning: Interpretation, Interpretability, Trustworthiness, and Beyondcited by 116
  11. 2024Inducing High Energy-Latency of Large Vision-Language Models with Verbose Imagescited by 69
  12. 2023OT-Attack: Enhancing Adversarial Transferability of Vision-Language Models via Optimal Transport Optimizationcited by 36
  13. 2023Analyzing Chain-of-Thought Prompting in Large Language Models via Gradient-based Feature Attributionscited by 32
  14. 2023Complex Query Answering on Eventuality Knowledge Graph with Implicit Logical Constraintscited by 21
  15. 2019Saliency Methods for Explaining Adversarial Attackscited by 16
  16. 2021Effective and Efficient Vote Attack on Capsule Networkscited by 14

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 5; 2025: 16; 2026: 5.

26 papers.

Sort:

By area:

  1. 2024A Survey of Attacks on Large Vision-Language Models: Resources, Advances, and Future Trendscited by 82
  2. 2025A Survey of Safety on Large Vision-Language Models: Attacks, Defenses and Evaluationscited by 66
  3. 2024Adversarial Robustness for Visual Grounding of Multimodal Large Language Modelscited by 23
  4. 2024Unveiling Typographic Deceptions: Insights of the Typographic Vulnerability in Large Vision-Language Modelcited by 22
  5. 2025X-Transfer Attacks: Towards Super Transferable Adversarial Attacks on CLIPcited by 21
  6. 2025Are Smarter LLMs Safer? Exploring Safety-Reasoning Trade-offs in Prompting and Fine-Tuningcited by 19
  7. 2024EAGLE: Elevating Geometric Reasoning through LLM-empowered Visual Instruction Tuningcited by 16
  8. 2025On the Robustness of GUI Grounding Models Against Image Attackscited by 14
  9. 2025From LLMs to MLLMs to Agents: A Survey of Emerging Paradigms in Jailbreak Attacks and Defenses within LLM Ecosystemcited by 11
  10. 2025AdInject: Real-World Black-Box Attacks on Web Agents via Advertising Deliverycited by 9
  11. 2025CHAI: Command Hijacking against embodied AIcited by 9
  12. 2025V-Attack: Targeting Disentangled Value Features for Controllable Adversarial Attacks on LVLMscited by 9
14 more
  1. 2024A Survey on Responsible Generative AI: What to Generate and What Notcited by 8
  2. 2025Knowing the Answer Isn't Enough: Fixing Reasoning Path Failures in LVLMscited by 5
  3. 2025Confidence Elicitation: A New Attack Vector for Large Language Modelscited by 4
  4. 2025Can Large Multimodal Models Actively Recognize Faulty Inputs? A Systematic Evaluation Framework of Their Input Scrutiny Abilitycited by 3
  5. 2026On the Adversarial Robustness of Multimodal LLM Judgescited by 3
  6. 2025Watch, Listen, Understand, Mislead: Tri-modal Adversarial Attacks on Short Videos for Content Appropriateness Evaluationcited by 2
  7. 2025Deep Research Agents Brings Deeper Harmcited by 2
  8. 2026Children's Intelligence Tests Pose Challenges for MLLMs? KidGym: A 2D Grid-Based Reasoning Benchmark for MLLMscited by 2
  9. 2025From Test-taking to Cognitive Scaffolding: A Pedagogical Diagnostic Benchmark for LLMs on English Standardized Testscited by 1
  10. 2025Standardization of Neuromuscular Reflex Analysis -- Role of Fine-Tuned Vision-Language Model Consortium and OpenAI gpt-oss Reasoning LLM Enabled Decision Support Systemcited by 1
  11. 2025Visual CoT Makes VLMs Smarter but More Fragilecited by 1
  12. 2026Don't Look at the Numbers: Visual Anchoring Bias and Layer-wise Representation in VLMscited by 0
  13. 2026On Adversarial Vulnerability of Vision-Language Models through the Lens of Intermediate Spectral Subspacescited by 0
  14. 2026Assessing Adversarial Robustness of Latent Reasoning Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026CAP-CoT: Cycle Adversarial Prompt for Improving Chain of Thoughts in LLM Reasoning
  2. 2026Are Reasoning LLMs Robust to Interventions on Their Chain-of-Thought?
  3. 2026Look Light, Think Heavy: What Multimodal Chain-of-Thought Reasoning Can and Cannot Do
  4. 2025MM-CoT:A Benchmark for Probing Visual Chain-of-Thought Reasoning in Multimodal Models
  5. 2025L2V-CoT: Cross-Modal Transfer of Chain-of-Thought Reasoning via Latent Intervention
  6. 2025From Perception to Reasoning: Deep Thinking Empowers Multimodal Large Language Models
  7. 2026TVI-CoT: Text-Visual Interleaved Chain-of-Thought Reasoning for Multimodal Understanding
  8. 2024Enhancing Adversarial Attacks through Chain of Thought