Guiding Diffusion-based Reconstruction with Contrastive Signals for Balanced Visual Representation

• URL: http://arxiv.org/abs/2603.04803v1
• Date: Thu, 05 Mar 2026 04:45:49 GMT
• Title: Guiding Diffusion-based Reconstruction with Contrastive Signals for Balanced Visual Representation
• Authors: Boyu Han, Qianqian Xu, Shilong Bao, Zhiyong Yang, Ruochen Cui, Xilin Zhao, Qingming Huang,
• Abstract summary: Contrastive Language-Image Pre-training (CLIP) has become a key bottleneck for downstream performance.<n>Recent solutions use diffusion models to enhance representations by conditioning image reconstruction on CLIP visual tokens.<n>We integrate contrastive signals into diffusion-based reconstruction to pursue more comprehensive visual representations.
• Score: 81.40978077888693
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The limited understanding capacity of the visual encoder in Contrastive Language-Image Pre-training (CLIP) has become a key bottleneck for downstream performance. This capacity includes both Discriminative Ability (D-Ability), which reflects class separability, and Detail Perceptual Ability (P-Ability), which focuses on fine-grained visual cues. Recent solutions use diffusion models to enhance representations by conditioning image reconstruction on CLIP visual tokens. We argue that such paradigms may compromise D-Ability and therefore fail to effectively address CLIP's representation limitations. To address this, we integrate contrastive signals into diffusion-based reconstruction to pursue more comprehensive visual representations. We begin with a straightforward design that augments the diffusion process with contrastive learning on input images. However, empirical results show that the naive combination suffers from gradient conflict and yields suboptimal performance. To balance the optimization, we introduce the Diffusion Contrastive Reconstruction (DCR), which unifies the learning objective. The key idea is to inject contrastive signals derived from each reconstructed image, rather than from the original input, into the diffusion process. Our theoretical analysis shows that the DCR loss can jointly optimize D-Ability and P-Ability. Extensive experiments across various benchmarks and multi-modal large language models validate the effectiveness of our method. The code is available at https://github.com/boyuh/DCR.

Related papers

• ReLAPSe: Reinforcement-Learning-trained Adversarial Prompt Search for Erased concepts in unlearned diffusion models [12.021923446217722]
  Machine unlearning is a key defense mechanism for removing unauthorized concepts from text-to-image diffusion models.<n>Existing adversarial approaches for exploiting this leakage are constrained by fundamental limitations.<n>We introduce ReLAPSe, a policy-based adversarial framework that reformulates concept restoration as a reinforcement learning problem.
  arXiv  Detail & Related papers  (2026-01-30T21:56:50Z)
• Eliminating Hallucination in Diffusion-Augmented Interactive Text-to-Image Retrieval [11.644783748459787]
  Diffusion-Augmented Interactive Text-to-Image Retrieval (DAI-TIR) is a promising paradigm that improves retrieval performance by generating query images via diffusion models.<n>We propose Diffusion-aware Multi-view Contrastive Learning (DMCL), a hallucination-robust training framework that casts DAI-TIR as joint optimization over representations of query intent and the target image.
  arXiv  Detail & Related papers  (2026-01-28T08:58:57Z)
• Diffusion-Augmented Contrastive Learning: A Noise-Robust Encoder for Biosignal Representations [0.4061135251278187]
  We propose a novel hybrid framework, Diffusion-Augmented Contrastive Learning (DACL), that fuses concepts from diffusion models and supervised contrastive learning.<n>It operates on a latent space created by a lightweight Variational Autoencoder (VAE) trained on our novel Scattering Transformer (ST) features.<n>A U-Net style encoder is then trained with a supervised contrastive objective to learn a representation that balances class discrimination with robustness to noise across various diffusion time steps.
  arXiv  Detail & Related papers  (2025-09-24T12:15:35Z)
• One-Step Diffusion-based Real-World Image Super-Resolution with Visual Perception Distillation [53.24542646616045]
  We propose VPD-SR, a novel visual perception diffusion distillation framework specifically designed for image super-resolution (SR) generation.<n>VPD-SR consists of two components: Explicit Semantic-aware Supervision (ESS) and High-frequency Perception (HFP) loss.<n>The proposed VPD-SR achieves superior performance compared to both previous state-of-the-art methods and the teacher model with just one-step sampling.
  arXiv  Detail & Related papers  (2025-06-03T08:28:13Z)
• "Principal Components" Enable A New Language of Images [79.45806370905775]
  We introduce a novel visual tokenization framework that embeds a provable PCA-like structure into the latent token space.<n>Our approach achieves state-of-the-art reconstruction performance and enables better interpretability to align with the human vision system.
  arXiv  Detail & Related papers  (2025-03-11T17:59:41Z)
• Semantic-Aligned Adversarial Evolution Triangle for High-Transferability Vision-Language Attack [51.16384207202798]
  Vision-language pre-training models are vulnerable to multimodal adversarial examples (AEs) Previous approaches augment image-text pairs to enhance diversity within the adversarial example generation process. We propose sampling from adversarial evolution triangles composed of clean, historical, and current adversarial examples to enhance adversarial diversity.
  arXiv  Detail & Related papers  (2024-11-04T23:07:51Z)
• Sample what you cant compress [4.950746640609865]
  We show how to learn a continuous encoder and decoder under a diffusion-based loss.<n>We demonstrate that this approach yields better reconstruction quality as compared to GAN-based autoencoders.<n>We also show that the resulting representation is easier to model with a latent diffusion model as compared to the representation obtained from a state-of-the-art GAN-based loss.
  arXiv  Detail & Related papers  (2024-09-04T08:42:42Z)
• Bridging Generative and Discriminative Models for Unified Visual Perception with Diffusion Priors [56.82596340418697]
  We propose a simple yet effective framework comprising a pre-trained Stable Diffusion (SD) model containing rich generative priors, a unified head (U-head) capable of integrating hierarchical representations, and an adapted expert providing discriminative priors. Comprehensive investigations unveil potential characteristics of Vermouth, such as varying granularity of perception concealed in latent variables at distinct time steps and various U-net stages. The promising results demonstrate the potential of diffusion models as formidable learners, establishing their significance in furnishing informative and robust visual representations.
  arXiv  Detail & Related papers  (2024-01-29T10:36:57Z)
• DiffBIR: Towards Blind Image Restoration with Generative Diffusion Prior [70.46245698746874]
  We present DiffBIR, a general restoration pipeline that could handle different blind image restoration tasks. DiffBIR decouples blind image restoration problem into two stages: 1) degradation removal: removing image-independent content; 2) information regeneration: generating the lost image content. In the first stage, we use restoration modules to remove degradations and obtain high-fidelity restored results. For the second stage, we propose IRControlNet that leverages the generative ability of latent diffusion models to generate realistic details.
  arXiv  Detail & Related papers  (2023-08-29T07:11:52Z)
• End-to-End Diffusion Latent Optimization Improves Classifier Guidance [81.27364542975235]
  Direct Optimization of Diffusion Latents (DOODL) is a novel guidance method. It enables plug-and-play guidance by optimizing diffusion latents. It outperforms one-step classifier guidance on computational and human evaluation metrics.
  arXiv  Detail & Related papers  (2023-03-23T22:43:52Z)
• Combining Reconstruction and Contrastive Methods for Multimodal Representations in RL [16.792949555151978]
  Learning self-supervised representations using reconstruction or contrastive losses improves performance and sample complexity of image-based and multimodal reinforcement learning (RL) Here, different self-supervised loss functions have distinct advantages and limitations depending on the information density of the underlying sensor modality. We propose Contrastive Reconstructive Aggregated representation Learning (CoRAL), a unified framework enabling us to choose the most appropriate self-supervised loss for each sensor modality.
  arXiv  Detail & Related papers  (2023-02-10T15:57:20Z)

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.