mlsanityData to 13 Oct 2026

ORCA: An Agentic Reasoning Framework for Hallucination and Adversarial Robustness in Vision-Language Models

arXiv 2509.15435 ↗First posted 18 Sep 2025cs.CV cs.AI cs.MAcited by 0

On the map: LLM reasoning → Hallucination detection and mitigation

also in 2 more
Abstract (arXiv)

Large Vision-Language Models (LVLMs) exhibit strong multimodal capabilities but remain vulnerable to hallucinations from intrinsic errors and adversarial attacks from external exploitations, limiting their reliability in real-world applications. We present ORCA, an agentic reasoning framework that improves the factual accuracy and adversarial robustness of pretrained LVLMs through inference-time structured inference reasoning with a suite of small vision models (less than 3B parameters). ORCA operates via an Observe-Reason-Critique-Act loop, querying multiple visual tools with evidential questions, validating cross-model inconsistencies, and refining predictions iteratively without access to model internals or retraining. ORCA also stores intermediate reasoning traces, which supports auditable decision-making. Though designed primarily to mitigate object-level hallucinations, ORCA also exhibits emergent adversarial robustness without requiring adversarial training or defense mechanisms. We evaluate ORCA across three settings: (1) clean images on hallucination benchmarks, (2) adversarially perturbed images without defense, and (3) adversarially perturbed images with defense applied. On the POPE hallucination benchmark, ORCA improves standalone LVLMs performance by +3.64% to +40.67% across different subsets. Under adversarial perturbations on POPE, ORCA achieves an average accuracy gain of +20.11% across LVLMs. When combined with defense techniques on adversarially perturbed AMBER images, ORCA further improves standalone LVLM performance, with gains ranging from +1.20% to +48.00% across metrics. These results demonstrate that ORCA offers a promising path toward building more reliable and robust multimodal systems.

Key facts. No citing papers in the map yet (data through 13 Oct 2026). Placed in Hallucination detection and mitigation, one of 17 directions in LLM reasoning. Builds on 33 of its 38 references that are in the map. First posted on arXiv 18 Sep 2025.

Builds on

33 of 38 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024Combating Multimodal LLM Hallucination via Bottom-Up Holistic Reasoning15 Dec

Other references · 32

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,588
  2. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,502
  3. 2020End-to-End Object Detection with Transformerscited by 7,396
  4. 2022BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generationcited by 4,314
  5. 2023MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Modelscited by 2,703
  6. 2023A Survey on Large Language Model based Autonomous Agentscited by 2,218
  7. 2023Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyondcited by 2,204
  8. 2023Evaluating Object Hallucination in Large Vision-Language Modelscited by 1,709
  9. 2023ImageBind: One Embedding Space To Bind Them Allcited by 1,168
  10. 2023mPLUG-Owl: Modularization Empowers Large Language Models with Multimodalitycited by 1,040
  11. 2023A Survey on Multimodal Large Language Modelscited by 905
  12. 2024PaliGemma: A versatile 3B VLM for transfercited by 686
20 more
  1. 2017Feature Squeezing: Detecting Adversarial Examples in Deep Neural Networkscited by 486
  2. 2024Hallucination of Multimodal Large Language Models: A Surveycited by 384
  3. 2024A Survey on Hallucination in Large Vision-Language Modelscited by 352
  4. 2023AMBER: An LLM-free Multi-dimensional Benchmark for MLLMs Hallucination Evaluationcited by 296
  5. 2023On Evaluating Adversarial Robustness of Large Vision-Language Modelscited by 284
  6. 2023Automatically Correcting Large Language Models: Surveying the landscape of diverse self-correction strategiescited by 237
  7. 2023Woodpecker: Hallucination Correction for Multimodal Large Language Modelscited by 226
  8. 2016A study of the effect of JPG compression on adversarial imagescited by 197
  9. 2023One Transformer Fits All Distributions in Multi-Modal Diffusion at Scalecited by 186
  10. 2024Robust CLIP: Unsupervised Adversarial Fine-Tuning of Vision Embeddings for Robust Large Vision-Language Modelscited by 120
  11. 2024A Survey of Attacks on Large Vision-Language Models: Resources, Advances, and Future Trendscited by 84
  12. 2024Critic-V: VLM Critics Help Catch VLM Errors in Multimodal Reasoningcited by 50
  13. 2024One Prompt Word is Enough to Boost Adversarial Robustness for Pre-trained Vision-Language Modelscited by 48
  14. 2024Logical Closed Loop: Uncovering Object Hallucinations in Large Vision-Language Modelscited by 47
  15. 2024Efficient Generation of Targeted and Transferable Adversarial Examples for Vision-Language Models Via Diffusion Modelscited by 38
  16. 2023Adversarial Illusions in Multi-Modal Embeddingscited by 29
  17. 2024Alleviating Hallucination in Large Vision-Language Models with Active Retrieval Augmentationcited by 29
  18. 2024Revisiting the Adversarial Robustness of Vision Language Models: a Multimodal Perspectivecited by 28
  19. 2024Exploiting Vision Encoder Vulnerabilities for Universal Adversarial Perturbations on Large Vision-Language Modelscited by 11
  20. 2024Seeing is Deceiving: Exploitation of Visual Pathways in Multi-Modal Language Modelscited by 4

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Hydra: An Agentic Reasoning Approach for Enhancing Adversarial Robustness and Mitigating Hallucinations in Vision-Language Models
  2. 2026HIVE: Understanding Post-Hallucination Reasoning in Vision Language Models
  3. 2024Towards a Systematic Evaluation of Hallucinations in Large-Vision Language Models
  4. 2025Visual hallucination detection in large vision-language models via evidential conflict
  5. 2026Vision-Language Introspection: Mitigating Overconfident Hallucinations in MLLMs via Interpretable Bi-Causal Steering
  6. 2024Pelican: Correcting Hallucination in Vision-LLMs via Claim Decomposition and Program of Thought Verification
  7. 2025On Epistemic Uncertainty of Visual Tokens for Object Hallucinations in Large Vision-Language Models
  8. 2024Reducing Hallucinations in Vision-Language Models via Latent Space Steering