S2H-DPO: Hardness-Aware Preference Optimization for Vision-Language Models

• URL: http://arxiv.org/abs/2604.18512v1
• Date: Mon, 20 Apr 2026 17:06:20 GMT
• Title: S2H-DPO: Hardness-Aware Preference Optimization for Vision-Language Models
• Abstract summary: We introduce a learning framework that constructs multi-image preference data across three hierarchical reasoning levels.<n>Our approach maintains strong single-image reasoning performance while simultaneously strengthening multi-image understanding capabilities.
• Score: 8.92478226038086
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Vision-Language Models (VLMs) have demonstrated remarkable progress in single-image understanding, yet effective reasoning across multiple images remains challenging. We identify a critical capability gap in existing multi-image alignment approaches: current methods focus primarily on localized reasoning with pre-specified image indices (``Look at Image 3 and...''), bypassing the essential skills of global visual search and autonomous cross-image comparison. To address this limitation, we introduce a Simple-to-Hard (S2H) learning framework that systematically constructs multi-image preference data across three hierarchical reasoning levels requiring an increasing level of capabilities: (1) single-image localized reasoning, (2) multi-image localized comparison, and (3) global visual search. Unlike prior work that relies on model-specific attributes, such as hallucinations or attention heuristics, to generate preference pairs, our approach leverages prompt-driven complexity to create chosen/rejected pairs that are applicable across different models. Through extensive evaluations on LLaVA and Qwen-VL models, we show that our diverse multi-image reasoning data significantly enhances multi-image reasoning performance, yielding significant improvements over baseline methods across benchmarks. Importantly, our approach maintains strong single-image reasoning performance while simultaneously strengthening multi-image understanding capabilities, thus advancing the state of the art for holistic visual preference alignment.

Related papers

• Learning Deep Modality-Shared Self-Expressiveness for Image Clustering with Textual Information [10.086588302174263]
  We propose a principled approach, termed deep modality-shared self-expressive model (DeepMORSE)<n>DeepMORSE discovers cross-modal structures via a modality-shared self-expressive model and simultaneously learns structured representations that conform to a union of modality-specific subspaces.<n>We evaluate our DeepMORSE on six widely used image clustering benchmarks and observe performance improvements exceeding 3%.
  arXiv  Detail & Related papers  (2026-08-09T02:23:49Z)
• Unified Multimodal Autoregressive Modeling with Shared Context-Visual Tokenizer is Key to Unification [80.62512020268626]
  UniAR is a unified autoregressive framework where a single visual tokenizer serves as the key bridge between understanding and generation.<n>UniAR adapts a pretrained vision encoder with multi-level feature fusion and a lookup-free bitwise quantization scheme.<n>A diffusion-based visual decoder operates on discrete visual tokens to decode high-fidelity images.
  arXiv  Detail & Related papers  (2026-06-16T17:59:22Z)
• Towards Generalized Multi-Image Editing for Unified Multimodal Models [56.620038824933566]
  Unified Multimodal Models (UMMs) integrate multimodal understanding and generation.<n>UMMs are limited to maintaining visual consistency and disambiguating visual cues when referencing details across multiple input images.<n>We propose a scalable multi-image editing framework for UMMs that explicitly distinguishes image identities and generalizes to variable input counts.
  arXiv  Detail & Related papers  (2026-01-09T06:42:49Z)
• Vision-Free Retrieval: Rethinking Multimodal Search with Textual Scene Descriptions [81.33113485830711]
  We introduce a vision-free, single-encoder retrieval pipeline for vision-language models.<n>We migrate to a text-to-text paradigm with the assistance of VLLM-generated structured image descriptions.<n>Our approach achieves state-of-the-art zero-shot performance on multiple retrieval and compositionality benchmarks.
  arXiv  Detail & Related papers  (2025-09-23T16:22:27Z)
• Simple o3: Towards Interleaved Vision-Language Reasoning [38.46230601239066]
  We propose Simple o3, an end-to-end framework that integrates dynamic tool interactions into interleaved vision-language reasoning.<n>Our approach features a scalable data synthesis pipeline that generates high-quality interleaved vision-language reasoning chains.<n> Experimental results demonstrate Simple o3's superior performance on diverse benchmarks, outperforming existing approaches.
  arXiv  Detail & Related papers  (2025-08-16T17:15:39Z)
• MiCo: Multi-image Contrast for Reinforcement Visual Reasoning [72.81576836419373]
  Chain-of-Thought (CoT) reasoning can be used to link visual cues across multiple images.<n>We adapt rule-based reinforcement learning for Vision-Language Models (VLMs)<n>Our method achieves significant improvements on multi-image reasoning benchmarks and shows strong performance on general vision tasks.
  arXiv  Detail & Related papers  (2025-06-27T17:59:27Z)
• PeRL: Permutation-Enhanced Reinforcement Learning for Interleaved Vision-Language Reasoning [50.21619363035618]
  We propose a general reinforcement learning approach PeRL tailored for interleaved multimodal tasks.<n>We introduce permutation of image sequences to simulate varied positional relationships to explore more spatial and positional diversity.<n>Our experiments confirm that PeRL trained model consistently surpasses R1-related and interleaved VLM baselines by a large margin.
  arXiv  Detail & Related papers  (2025-06-17T18:25:56Z)
• Weaving Context Across Images: Improving Vision-Language Models through Focus-Centric Visual Chains [31.828341309787042]
  Vision-language models (VLMs) achieve remarkable success in single-image tasks.<n>Real-world scenarios often involve intricate multi-image inputs, leading to a notable performance decline.<n>We propose Focus-Centric Visual Chain, a novel paradigm that enhances VLMs'perception, comprehension, and reasoning abilities in multi-image scenarios.
  arXiv  Detail & Related papers  (2025-04-28T19:02:18Z)
• Unified Reward Model for Multimodal Understanding and Generation [32.22714522329413]
  This paper proposes UnifiedReward, the first unified reward model for multimodal understanding and generation assessment.<n>We first develop UnifiedReward on our constructed large-scale human preference dataset, including both image and video generation/understanding tasks.
  arXiv  Detail & Related papers  (2025-03-07T08:36:05Z)
• ImageChain: Advancing Sequential Image-to-Text Reasoning in Multimodal Large Language Models [12.265270657795275]
  ImageChain is a framework that enhances MLLMs with sequential reasoning capabilities over image data.<n>Our approach improves performance on the next-scene description task.<n>ImageChain achieves robust zero-shot out-of-domain performance in applications ranging from comics to robotics.
  arXiv  Detail & Related papers  (2025-02-26T18:55:06Z)
• MIA-DPO: Multi-Image Augmented Direct Preference Optimization For Large Vision-Language Models [85.30735602813093]
  Multi-Image Augmented Direct Preference Optimization (MIA-DPO) is a visual preference alignment approach that effectively handles multi-image inputs. MIA-DPO mitigates the scarcity of diverse multi-image training data by extending single-image data with unrelated images arranged in grid collages or pic-in-pic formats.
  arXiv  Detail & Related papers  (2024-10-23T07:56:48Z)
• Matryoshka Multimodal Models [92.41824727506751]
  We propose M3: Matryoshka Multimodal Models, which learns to represent visual content as nested sets of visual tokens. We find that COCO-style benchmarks only need around 9 visual tokens to obtain accuracy similar to that of using all 576 tokens.
  arXiv  Detail & Related papers  (2024-05-27T17:59:56Z)

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.