Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
131 tokens/sec
GPT-4o
10 tokens/sec
Gemini 2.5 Pro Pro
47 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Mitigating Object Hallucination in MLLMs via Data-augmented Phrase-level Alignment (2405.18654v3)

Published 28 May 2024 in cs.CV

Abstract: Despite their significant advancements, Multimodal LLMs (MLLMs) often generate factually inaccurate information, referred to as hallucination. In this work, we address object hallucinations in MLLMs, where information is generated about an object not present in the input image. We introduce Data-augmented Phrase-level Alignment (DPA), a novel loss which can be applied to instruction-tuned off-the-shelf MLLMs to mitigate hallucinations, while preserving their general vision-language capabilities. To fine-tune MLLMs with DPA, we first generate a set of hallucinated' andcorrect' response pairs through generative data augmentation by selectively altering the ground-truth information of the correct responses at a phrase level. The DPA loss is then used to train MLLMs to reduce the likelihood of hallucinated phrases compared to the correct ones. Our thorough evaluation on various benchmarks confirms the effectiveness of DPA in mitigating hallucination while retaining the out-of-the-box performance of the MLLMs on general tasks. For instance, MLLMs finetuned with DPA, which we refer to as Hallucination Attenuated Language and Vision Assistant (HALVA), improve F1 by up to 13.4% on hallucination visual question-answering and reduce the hallucination rate by up to 4.2% on image description tasks.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (78)
  1. Palm: Scaling language modeling with pathways. JMLR, 24(240):1–113, 2023.
  2. Palm 2 technical report. arXiv preprint arXiv:2305.10403, 2023.
  3. Exploring the limits of transfer learning with a unified text-to-text transformer. JMLR, 21(140):1–67, 2020.
  4. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023.
  5. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023.
  6. Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805, 2023.
  7. Language models are few-shot learners. NeurIPS, 33:1877–1901, 2020.
  8. Visual instruction tuning. NeurIPS, 36, 2024.
  9. Improved baselines with visual instruction tuning. arXiv preprint arXiv:2310.03744, 2023.
  10. Instructblip: Towards general-purpose vision-language models with instruction tuning. arXiv preprint arXiv:2305.06500, 2023.
  11. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. arXiv preprint arXiv:2301.12597, 2023.
  12. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023.
  13. Chatcad: Interactive computer-aided diagnosis on medical image using large language models. arXiv preprint arXiv:2302.07257, 2023.
  14. Advancing medical imaging with language models: A journey from n-grams to chatgpt. arXiv preprint arXiv:2304.04920, 2023.
  15. Object hallucination in image captioning. arXiv preprint arXiv:1809.02156, 2018.
  16. Hallucination of multimodal large language models: A survey. arXiv preprint arXiv:2404.18930, 2024.
  17. Mitigating hallucination in large multi-modal models via robust instruction tuning. In ICLR, 2023.
  18. Seeing is believing: Mitigating hallucination in large vision-language models via clip-guided decoding. arXiv preprint arXiv:2402.15300, 2024.
  19. Woodpecker: Hallucination correction for multimodal large language models. arXiv preprint arXiv:2310.16045, 2023.
  20. Mitigating object hallucinations in large vision-language models through visual contrastive decoding. arXiv preprint arXiv:2311.16922, 2023.
  21. Volcano: mitigating multimodal hallucination through self-feedback guided revision. arXiv preprint arXiv:2311.07362, 2023.
  22. Analyzing and mitigating object hallucination in large vision-language models. arXiv preprint arXiv:2310.00754, 2023.
  23. Let there be a clock on the beach: Reducing object hallucination in image captioning. In WACV, pages 1381–1390, 2022.
  24. Aligning large multimodal models with factually augmented rlhf. arXiv preprint arXiv:2309.14525, 2023.
  25. Hallucination augmented contrastive learning for multimodal large language model. arXiv preprint arXiv:2312.06968, 2023.
  26. Beyond hallucinations: Enhancing lvlms through hallucination-aware direct preference optimization. arXiv preprint arXiv:2311.16839, 2023.
  27. Less is more: Mitigating multimodal hallucination from an eos decision perspective. arXiv preprint arXiv:2402.14545, 2024.
  28. Ibd: Alleviating hallucinations in large vision-language models via image-biased decoding. arXiv preprint arXiv:2402.18476, 2024.
  29. Logical closed loop: Uncovering object hallucinations in large vision-language models. arXiv preprint arXiv:2402.11622, 2024.
  30. Rlhf-v: Towards trustworthy mllms via behavior alignment from fine-grained correctional human feedback. arXiv preprint arXiv:2312.00849, 2023.
  31. Mocha: Multi-objective reinforcement mitigating caption hallucinations. arXiv preprint arXiv:2312.03631, 2023.
  32. Understanding and improving robustness of vision transformers through patch-based negative augmentation. NeurIPS, 35:16276–16289, 2022.
  33. Toward understanding generative data augmentation. NeurIPS, 36, 2024.
  34. Deep reinforcement learning from human preferences. NeurIPS, 30, 2017.
  35. Direct preference optimization: Your language model is secretly a reward model. NeurIPS, 36, 2024.
  36. Jonathon Shlens. Notes on kullback-leibler divergence and likelihood. arXiv preprint arXiv:1404.2000, 2014.
  37. Evaluating object hallucination in large vision-language models. arXiv preprint arXiv:2305.10355, 2023.
  38. Neural machine translation of rare words with subword units. arXiv preprint arXiv:1508.07909, 2015.
  39. Sentencepiece: A simple and language independent subword tokenizer and detokenizer for neural text processing. arXiv preprint arXiv:1808.06226, 2018.
  40. Visual genome: Connecting language and vision using crowdsourced dense image annotations. IJCV, 123:32–73, 2017.
  41. Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality, March 2023.
  42. Learning transferable visual models from natural language supervision. In ICML, pages 8748–8763. PMLR, 2021.
  43. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685, 2021.
  44. Mme: A comprehensive evaluation benchmark for multimodal large language models. arXiv preprint arXiv:2306.13394, 2023.
  45. An llm-free multi-dimensional benchmark for mllms hallucination evaluation. arXiv preprint arXiv:2311.07397, 2023.
  46. Hallusionbench: You see what you think? or you think what you see? an image-context reasoning benchmark challenging for gpt-4v (ision), llava-1.5, and other multi-modality models. arXiv preprint arXiv:2310.14566, 2023.
  47. Making the v in vqa matter: Elevating the role of image understanding in visual question answering. In CVPR, pages 6904–6913, 2017.
  48. Mm-vet: Evaluating large multimodal models for integrated capabilities. arXiv preprint arXiv:2308.02490, 2023.
  49. Towards vqa models that can read. In CVPR, pages 8317–8326, 2019.
  50. Opera: Alleviating hallucination in multi-modal large language models via over-trust penalty and retrospection-allocation. arXiv preprint arXiv:2311.17911, 2023.
  51. Minigpt-v2: large language model as a unified interface for vision-language multi-task learning. arXiv preprint arXiv:2310.09478, 2023.
  52. mplug-owl: Modularization empowers large language models with multimodality. arXiv preprint arXiv:2304.14178, 2023.
  53. Multimodal-gpt: A vision and language model for dialogue with humans. arXiv preprint arXiv:2305.04790, 2023.
  54. Minigpt-4: Enhancing vision-language understanding with advanced large language models. arXiv preprint arXiv:2304.10592, 2023.
  55. Empowering vision-language models to follow interleaved vision-language instructions. arXiv preprint arXiv:2308.04152, 2023.
  56. Mimic-it: Multi-modal in-context instruction tuning. arXiv preprint arXiv:2306.05425, 2023.
  57. mplug-owl2: Revolutionizing multi-modal large language model with modality collaboration. arXiv preprint arXiv:2311.04257, 2023.
  58. What matters in training a gpt4-style language model with multimodal inputs? arXiv preprint arXiv:2307.02469, 2023.
  59. Qwen-vl: A versatile vision-language model for understanding, localization, text reading, and beyond. 2023.
  60. Llama-adapter v2: Parameter-efficient visual instruction model. arXiv preprint arXiv:2304.15010, 2023.
  61. Bliva: A simple multimodal llm for better handling of text-rich visual questions. In AAAI, volume 38, pages 2256–2264, 2024.
  62. Mmicl: Empowering vision-language model with multi-modal in-context learning. arXiv preprint arXiv:2309.07915, 2023.
  63. Sphinx: The joint mixing of weights, tasks, and visual embeddings for multi-modal large language models. arXiv preprint arXiv:2311.07575, 2023.
  64. Muffin: Curating multi-faceted instructions for improving instruction following. In ICLR, 2023.
  65. Kosmos-2: Grounding multimodal large language models to the world. arXiv preprint arXiv:2306.14824, 2023.
  66. Obelics: An open web-scale filtered dataset of interleaved image-text documents. NeurIPS, 36, 2024.
  67. Minigpt-5: Interleaved vision-and-language generation via generative vokens. arXiv preprint arXiv:2310.02239, 2023.
  68. Openflamingo: An open-source framework for training large autoregressive vision-language models. arXiv preprint arXiv:2308.01390, 2023.
  69. Scaling up visual and vision-language representation learning with noisy text supervision. In ICML, pages 4904–4916. PMLR, 2021.
  70. Coca: Contrastive captioners are image-text foundation models. arXiv preprint arXiv:2205.01917, 2022.
  71. Pali: A jointly-scaled multilingual language-image model. arXiv preprint arXiv:2209.06794, 2022.
  72. Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation. In ICML, pages 12888–12900. PMLR, 2022.
  73. Ofa: Unifying architectures, tasks, and modalities through a simple sequence-to-sequence learning framework. In ICML, pages 23318–23340. PMLR, 2022.
  74. Shikra: Unleashing multimodal llm’s referential dialogue magic. arXiv preprint arXiv:2306.15195, 2023.
  75. Vigc: Visual instruction generation and correction. In AAAI, volume 38, pages 5309–5317, 2024.
  76. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2017.
  77. {{\{{Zero-offload}}\}}: Democratizing {{\{{billion-scale}}\}} model training. In USENIX ATC, pages 551–564, 2021.
  78. Zero-infinity: Breaking the gpu memory wall for extreme scale deep learning. In SC, pages 1–14, 2021.
Citations (3)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now
X Twitter Logo Streamline Icon: https://streamlinehq.com

Tweets

Youtube Logo Streamline Icon: https://streamlinehq.com

YouTube