Fighting Hallucinations with Counterfactuals: Diffusion-Guided Perturbations for LVLM Hallucination Suppression

• URL: http://arxiv.org/abs/2603.10470v1
• Date: Wed, 11 Mar 2026 06:40:50 GMT
• Title: Fighting Hallucinations with Counterfactuals: Diffusion-Guided Perturbations for LVLM Hallucination Suppression
• Abstract summary: We introduce CIPHER, a training-free method that suppresses vision-induced hallucinations via lightweight feature-level correction.<n>Unlike prior training-free approaches that primarily focus on text-induced hallucinations, CIPHER explicitly targets hallucinations arising from the visual modality.<n> Experiments across multiple benchmarks show that CIPHER significantly reduces hallucination rates while preserving task performance.
• Score: 3.2584778174972295
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: While large vision-language models (LVLMs) achieve strong performance on multimodal tasks, they frequently generate hallucinations -- unfaithful outputs misaligned with the visual input. To address this issue, we introduce CIPHER (Counterfactual Image Perturbations for Hallucination Extraction and Removal), a training-free method that suppresses vision-induced hallucinations via lightweight feature-level correction. Unlike prior training-free approaches that primarily focus on text-induced hallucinations, CIPHER explicitly targets hallucinations arising from the visual modality. CIPHER operates in two phases. In the offline phase, we construct OHC-25K (Object-Hallucinated Counterfactuals, 25,000 samples), a counterfactual dataset consisting of diffusion-edited images that intentionally contradict the original ground-truth captions. We pair these edited images with the unchanged ground-truth captions and process them through an LVLM to extract hallucination-related representations. Contrasting these representations with those from authentic (image, caption) pairs reveals structured, systematic shifts spanning a low-rank subspace characterizing vision-induced hallucination. In the inference phase, CIPHER suppresses hallucinations by projecting intermediate hidden states away from this subspace. Experiments across multiple benchmarks show that CIPHER significantly reduces hallucination rates while preserving task performance, demonstrating the effectiveness of counterfactual visual perturbations for improving LVLM faithfulness. Code and additional materials are available at https://hamidreza-dastmalchi.github.io/cipher-cvpr2026/.

Related papers

• When Prompts Override Vision: Prompt-Induced Hallucinations in LVLMs [54.411658510110215]
  We propose HalluScope to better understand the extent to which different factors induce hallucinations.<n>HalluVL-DPO is a framework for fine-tuning off-the-shelf LVLMs towards more visually grounded responses.
  arXiv  Detail & Related papers  (2026-04-23T17:54:36Z)
• CRoPS: A Training-Free Hallucination Mitigation Framework for Vision-Language Models [3.0991186209192794]
  Hallucinated content undermines reliability in real-world use.<n>New model captures hallucination effects by selectively removing key text tokens.<n>Generalized Contrastive Decoding integrates multiple hallucinated models to represent diverse hallucination sources.
  arXiv  Detail & Related papers  (2026-01-02T11:39:00Z)
• Mitigating Object and Action Hallucinations in Multimodal LLMs via Self-Augmented Contrastive Alignment [40.48528326378281]
  We propose a framework for enabling object and action faithfulness in dynamic videos.<n>SANTA employs a hallucinative self-augmentation scheme to identify the potential hallucinations.<n>We develop a tracklet-phrase contrastive alignment to match the regional objects and relation-guided actions with their corresponding visual and temporal phrases.
  arXiv  Detail & Related papers  (2025-12-04T01:05:16Z)
• Exposing Hallucinations To Suppress Them: VLMs Representation Editing With Generative Anchors [8.089908150148554]
  Multimodal large language models (MLLMs) have achieved remarkable success across diverse vision-language tasks.<n>MLLMs are highly susceptible to hallucinations, producing content that is fluent but inconsistent with visual evidence.<n>We propose a training-free, self-supervised method for hallucination mitigation.
  arXiv  Detail & Related papers  (2025-09-26T07:24:28Z)
• HELPD: Mitigating Hallucination of LVLMs by Hierarchical Feedback Learning with Vision-enhanced Penalty Decoding [36.360171373963716]
  Large Vision-Language Models (LVLMs) have shown remarkable performance on many visual-language tasks. These models still suffer from multimodal hallucination, which means the generation of objects or content that violates the images. We propose Hierarchical Feedback Learning with Vision-enhanced Penalty Decoding (HELPD) to address this issue.
  arXiv  Detail & Related papers  (2024-09-30T15:52:05Z)
• Does Object Grounding Really Reduce Hallucination of Large Vision-Language Models? [53.89380284760555]
  Large vision-language models (LVLMs) produce captions that mention concepts that cannot be found in the image. These hallucinations erode the trustworthiness of LVLMs and are arguably among the main obstacles to their ubiquitous adoption. Recent work suggests that addition of grounding objectives -- those that explicitly align image regions or objects to text spans -- reduces the amount of LVLM hallucination.
  arXiv  Detail & Related papers  (2024-06-20T16:56:11Z)
• Visual Description Grounding Reduces Hallucinations and Boosts Reasoning in LVLMs [52.497823009176074]
  Large Vision-Language Models (LVLMs) often produce responses that misalign with factual information, a phenomenon known as hallucinations.<n>We introduce Visual Description Grounded Decoding (VDGD), a training-free method designed to enhance visual perception and improve reasoning capabilities in LVLMs.
  arXiv  Detail & Related papers  (2024-05-24T16:21:59Z)
• Alleviating Hallucinations in Large Vision-Language Models through Hallucination-Induced Optimization [123.54980913741828]
  Large Visual Language Models (LVLMs) have demonstrated exceptional abilities in understanding multimodal data.<n>They invariably suffer from hallucinations, leading to a disconnect between the generated text and the corresponding images.<n>Almost all current visual contrastive decoding methods attempt to mitigate these hallucinations by introducing visual uncertainty information.<n>However, they struggle to precisely induce the hallucinatory tokens, which severely limits their effectiveness in mitigating hallucinations.
  arXiv  Detail & Related papers  (2024-05-24T08:46:31Z)
• ESREAL: Exploiting Semantic Reconstruction to Mitigate Hallucinations in Vision-Language Models [6.014286500397164]
  Hallucinations in vision-language models pose a significant challenge to their reliability, particularly in the generation of long captions. We introduce ESREAL, a novel unsupervised learning framework designed to suppress the generation of hallucinations through accurate localization and penalization of hallucinated tokens. Our framework notably reduces hallucinations in LLaVA, InstructBLIP, and mPLUG-Owl2 by 32.81%, 27.08%, and 7.46% on the CHAIR metric.
  arXiv  Detail & Related papers  (2024-03-24T14:21:06Z)
• Hallucination Augmented Contrastive Learning for Multimodal Large Language Model [53.65682783591723]
  Multi-modal large language models (MLLMs) have been shown to efficiently integrate natural language with visual information to handle multi-modal tasks. However, MLLMs still face a fundamental limitation of hallucinations, where they tend to generate erroneous or fabricated information. In this paper, we address hallucinations in MLLMs from a novel perspective of representation learning.
  arXiv  Detail & Related papers  (2023-12-12T04:05:15Z)
• Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding [125.05295513481035]
  We introduce Visual Contrastive Decoding (VCD), a simple and training-free method that contrasts output distributions derived from original and distorted visual inputs. The proposed VCD effectively reduces the over-reliance on statistical bias and unimodal priors, two essential causes of object hallucinations. Our experiments show that VCD, without either additional training or the usage of external tools, significantly mitigates the object hallucination issue across different LVLM families.
  arXiv  Detail & Related papers  (2023-11-28T16:26:35Z)
• HalluciDoctor: Mitigating Hallucinatory Toxicity in Visual Instruction Data [102.56792377624927]
  hallucinations inherent in machine-generated data remain under-explored. We present a novel hallucination detection and elimination framework, HalluciDoctor, based on the cross-checking paradigm. Our method successfully mitigates 44.6% hallucinations relatively and maintains competitive performance compared to LLaVA.
  arXiv  Detail & Related papers  (2023-11-22T04:52:58Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.