Skip to content

Latest commit

 

History

History
154 lines (139 loc) · 17.8 KB

File metadata and controls

154 lines (139 loc) · 17.8 KB

Large MultiModal Model Hallucination Awesome

LMM hallucination😵 refers to occasional instances where LMMs generate content that appears plausible but deviates from or conflicts with the provided image. LMMs tend to rely more on their own parametric knowledge than on provided visual features, causing them to respond with guesses and generate multimodal hallucinations.

In the MLLM community, we've developed methods for detecting, evaluating, and mitigating hallucinations👍.



Detecting

  1. FDPO: Detecting and Preventing Hallucinations in Large Vision Language Models, (Gunjal et al. 2023 AAAI 2024)
    • Static Badge Static Badge
  2. HaELM: Evaluation and Analysis of Hallucination in Large Vision-Language Models, (Wang et al. 2023a)
    • Static Badge
    • An automatic MLLM hallucination detection framework, Train LLM to detect
  3. HallE-Switch: Rethinking and Controlling Object Existence Hallucinations in Large Vision-Language Models for Detailed Caption, (Zhai et al. 2023)
    • Static Badge
  4. Unified Hallucination Detection for Multimodal Large Language Models, (Chen et al.)
    • Static Badge

Evaluating

  1. POPE: Evaluating Object Hallucination in Large Vision-Language Models, (Li et al. EMNLP 2023)
    • Static Badge
    • Discriminative Task: Object Existence, 3k * 3 VQA pairs
    • LLM-free
  2. HaELM: Evaluation and Analysis of Hallucination in Large Vision-Language Models, (Wang et al. 2023a)
    • Static Badge
    • Discriminative Task, 1500 VQA pairs
  3. HallusionBench: An Image-Context Reasoning Benchmark Challenging for GPT-4V(ision), LLaVA-1.5, and Other Multi-modality Model, (Liu et al. 2023, CVPR2024)
    • Static Badge
    • Image Reasoning Task, 200 VQA pairs
  4. NOPE: Negative Object Presence Evaluation (NOPE) to Measure Object Hallucination in Vision-Language Models, (Lovenia et al.)
    • Static Badge Static Badge
  5. Bingo: Holistic Analysis of Hallucination in GPT-4V(ision): Bias and Interference Challenges, (Cui et al.)
    • Static Badge
  6. FaithScore: Evaluating Hallucinations in Large Vision-Language Models, (Jing et al.)
    • Static Badge
    • Generative Task: Object Existence, Attribute, Relationship, 180 VQA pairs
    • open-end find-grained evaluation, need other models to help evaluation
  7. AMBER: An LLM-free Multi-dimensional Benchmark for MLLMs Hallucination Evaluation, (Wang et al.)
    • Static Badge
    • Discriminative Task: Object Existence, Attribute, Relationship
    • Generative Task: Object Existence
    • LLM-free
  8. Behind the Magic, MERLIM: Multi-modal Evaluation Benchmark for Large Image-Language Models, (Villa et al.)
    • Static Badge
  9. Visually Dehallucinative Instruction Generation: Know What You Don't Know, (Cha et al.)
    • Static Badge Static Badge
  10. The Instinctive Bias: Spurious Images lead to Hallucination in MLLMs, (Han et al.)
    • Static Badge
  11. Hal-Eval: A Universal and Fine-grained Hallucination Evaluation Framework for Large Vision Language Models, (Jiang et al.)
    • Static Badge Static Badge
  12. Visual Hallucinations of Multi-modal Large Language Models, (Huang et al.)
    • Static Badge
  13. PhD: A Prompted Visual Hallucination Evaluation Dataset, (Liu et al.)
    • Static Badge

Mitigating

  1. LRV-Instruction: Mitigating Hallucination in Large Multi-Modal Models via Robust Instruction Tuning, (Liu et al. ICLR2024)
    • Static Badge
    • [dataset] propose an instruction-tuning dataset that includes both positive and negative sample
    • GAIVE: evaluation approach which uses GPT-4
  2. LURE: Analyzing and Mitigating Object Hallucination in Large Vision-Language Models, (Zhou et al. ICLR2024)
    • Static Badge
    • [post-hoc revision] train a revision model to detect and correct hallucinated objects in the base model’s response.
  3. HallE-Switch: Rethinking and Controlling Object Existence Hallucinations in Large Vision-Language Models for Detailed Caption, (Zhai et al. 2023)
    • Static Badge
    • CCEval, a GPT-4 assisted evaluation method tailored for detailed captioning
  4. Woodpecker: Hallucination Correction for Multimodal Large Language Models, (Yin et al.)
    • Static Badge Static Badge
    • [revision] post-hoc correction
    • need other pre-trained visual models
  5. LLaVA-RLHF: Aligning Large Multimodal Models with Factually Augmented RLHF, (Sun et al.)
    • Static Badge
    • [RLHF-PPO] the first LMM trained with RLHF
    • propose benchmark: MMHal-Bench
  6. Volcano: Mitigating Multimodal Hallucination through Self-Feedback Guided Revision, (Lee et al., NAACL 2024)
    • Static Badge
    • self-feedback, according to self-generate natural language feedback to self-revise response
  7. HalluciDoctor: Mitigating Hallucinatory Toxicity in Visual Instruction Data, (Yu et al., CVPR2024)
    • Static Badge
  8. VCD: Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding, (Leng et al., CVPR 2024)
    • Static Badge
    • train-free
  9. HA-DPO: Beyond Hallucinations: Enhancing LVLMs through Hallucination-Aware Direct Preference Optimization
    • Static Badge
  10. Mitigating Hallucination in Visual Language Models with Visual Supervision, (Chen et al.)
    • Static Badge Static Badge
    • construct a fine-grained vision instruction dataset, RAI-30k. It contains multi-modal conversations focusing on specific vision relations in an image.
    • propose a new benchmark: RAHBench
    • incorporating SAM in the vision instruction tuning process'
  11. OPERA: Alleviating Hallucination in Multi-Modal Large Language Models via Over-Trust Penalty and Retrospection-Allocation, (Huang et al. CVPR 2024)
    • Static Badge
  12. FOHE: Mitigating Fine-Grained Hallucination by Fine-Tuning Large Vision-Language Models with Caption Rewrites, (Wang et al.)
    • Static Badge
    • use ChatGPT to post-hoc correction
  13. RLHF-V: Towards Trustworthy MLLMs via Behavior Alignment from Fine-grained Correctional Human Feedback, (CVPR 2024)
    • Static Badge
    • [RLHF-DPO] 1.4K preference data, natural language feedback
  14. MOCHa: Multi-Objective Reinforcement Mitigating Caption Hallucinations, (Ben-Kish et al.)
    • Static Badge
    • [RLHF]
  15. HACL: Hallucination Augmented Contrastive Learning for Multimodal Large Language Model, (Jiang et al.)
    • Static Badge Static Badge
  16. Silkie: Preference Distillation for Large Visual Language Models, (Li et al.)
    • Static Badge
  17. MMCot: Multimodal Chain-of-Thought Reasoning in Language Models, (Zhang et al.)
    • Static Badge
    • [CoT]
  18. KAM-CoT: Knowledge Augmented Multimodal Chain-of-Thoughts Reasoning, (Mondal et al. AAAI 2024)
    • Static Badge Static Badge
    • [CoT]
  19. Mitigating Object Hallucination in Large Vision-Language Models via Classifier-Free Guidance, (Zhao et al.)
    • Static Badge Static Badge
  20. ViGoR: Improving Visual Grounding of Large Vision Language Models with Fine-Grained Reward Modeling, (Yan et al.)
    • Static Badge Static Badge
  21. EFUF: Efficient Fine-grained Unlearning Framework for Mitigating Hallucinations in Multimodal Large Language Models, (Xing et al.)
    • Static Badge Static Badge
  22. Skip \n: A Simple Method to Reduce Hallucination in Large Vision-Language Models, (Han et al.)
    • Static Badge Static Badge
  23. Less is More: Mitigating Multimodal Hallucination from an EOS Decision Perspective, (Yue et al.)
    • Static Badge
    • overly detailed training data can lead to model output beyond visual perception limits, thus exhibiting hallucinations
    • propose a learning objective that reduces hallucinations by learning from regular instruction data
    • propose a data filtering strategy that prevents harmful training data from exacerbating model hallucinations
  24. Seeing is Believing: Mitigating Hallucination in Large Vision-Language Models via CLIP-Guided Decoding, (Deng et al.)
    • Static Badge Static Badge
  25. Navigating Hallucinations for Reasoning of Unintentional Activities, (Grover et al.)
    • Static Badge Static Badge
  26. IBD: Alleviating Hallucinations in Large Vision-Language Models via Image-Biased Decoding, (Zhu et al.)
    • Static Badge Static Badge
  27. Pensieve: Retrospect-then-Compare mitigates Visual Hallucination, (Yang et al.)
    • Static Badge
  28. M3ID: Multi-Modal Hallucination Control by Visual Information Grounding, (Favero et al. CVPR 2024)
    • Static Badge Static Badge
  29. What if...?: Counterfactual Inception to Mitigate Hallucination Effects in Large Multimodal Models, (Kim et al.)
    • Static Badge
  30. Mitigating Dialogue Hallucination for Large Multi-modal Models via Adversarial Instruction Tuning, (Park et al.)
    • Static Badge Static Badge