Discriminative models are trained with task-specific objectives such as classification, detection, or representation learning, yet their practical utility depends on generalizing beyond the training task to unseen samples from the same domain. This suggests that a discriminative model may retain an implicit trace of the data manifold beyond its task labels. We ask whether this trace can be exposed as a generative signal without task-specific guidance. We introduce the Discriminative Score Function (DSF), which extracts an image-space generative field from the loss-gradient geometry of a pretrained discriminative model. DSF provides a training-free update rule that moves images toward model-consistent visual regions without retraining or task-specific targets. We apply DSF to diverse pretrained discriminative models, including ResNet classifiers, DETR detectors, and DINOv2 ViT encoders. DSF generates unconditional images from noise and supports conditional guidance, editing, inpainting, and explanation through the same field. DSF-generated samples also provide effective zero-shot quantization calibration when real images or labels are unavailable, preserving feature geometry better than data-free calibration baselines. These results show that discriminative training leaves a generative functional signal for synthesis and calibration. Project page: https://junhoo.me/dsf.
Figures & tables
Figure 1: DSF as a functional generative prior. A candidate image is mapped through a fixed discriminative model into a loss-gradient functional space; DSF then provides an update direction that supports unconditional generation, conditional guidance, editing, inpainting, and explanation through the same interface.
Figure 2: Training maps nearby datasets on the same manifold to nearby trained functions.
Figure 3: Examples of unconditionally generated images using various discriminative models. Top: DINOv2/LVD-124M (self-supervised learning). Middle: DeTR/COCO (object detection). Bottom: ResNet/Imagenet (image classification).
Figure 4: Examples of conditionally generated images using DSF. Top: Both the score model and the conditional model used ResNet/Imagenet. Middle: The score model used DINOv2/LVD-124M, and the conditional model used ResNet/Imagenet. Bottom: The score model used ResNet/Imagenet, and the conditional model used CLIP.
Downstream usability
Representation geometry
Calibration
k-NN ↑
Head ↑
fpQ@5 ↑
CLS ↑
Patch ↑
8
6
4
8
6
4
8
6
4
8
6
4
8
6
4
Real images †
.659
.605
.580
.269
.195
.187
.882
.873
.823
.939
.798
.708
.968
.874
.753
DSF (ours)
.659
.605
.589
.283
.215
.202
.883
.872
.825
.947
.814
.710
.973
.885
.760
RegCache
.629
.564
.500
.196
.148
.142
.874
.859
.787
.902
.744
.608
.963
.849
.671
Gaussian
.592
.435
.077
.152
.051
.003
.874
.791
.058
.825
.515
.102
.913
.622
.153
Table 1: Zero-shot quantization with DSF calibration. DSF matches or exceeds the real-image oracle in nearly all settings and consistently outperforms data-free calibration sources for a quantized DINOv2 feature extractor. All sources use 100 calibration images and the same scale-selection protocol; higher is better.
Figure 5: Examples of editing through DSF. Top: original images; bottom:he top row contains original images while the bottom row shows edited images.
Figure 6: Examples of the inpainting task. For each set, the left is the inpainted image, the middle is the reconstructed one, and the right is the original image.
Figure 8
Appendix figures & tables6 assets
Supplementary material from the paper’s appendix.
Appendix
Asset
Use in this paper
License or terms followed
ResNet / torchvision [ 19 ]
Public pretrained classifier used as a DSF score model
PyTorch/torchvision code and model access are used under the PyTorch/torchvision BSD-style license terms; ImageNet-trained weights are used only for research evaluation.
DINOv2 [ 41 ]
Public pretrained self-supervised ViT used for generation and quantization diagnostics
DINOv2 code and model weights are released under Apache-2.0; pretrained checkpoints are used as fixed evaluation models and are not redistributed.
DETR / COCO [ 3 , 35 ]
Public pretrained detector used as a DSF score model
DETR is released under Apache-2.0; COCO data are used or referenced under the official COCO terms and associated image licenses.
ImageNet [ 9 ]
Oracle calibration images and downstream evaluation reference
Accessed under ImageNet terms for non-commercial research and educational use; images are not redistributed.
CLIP / OpenCLIP [ 6 ]
External guidance model for conditional generation
Public CLIP/OpenCLIP implementations and checkpoints are used under their respective repository/model licenses, including MIT-licensed OpenAI CLIP code, and are not redistributed.
SmoothQuant [ 48 ]
Post-training quantization transform used in zero-shot calibration diagnostics
SmoothQuant code is released under the MIT license.
Appendix
Table 2: Existing assets used or referenced in the experiments.
ℓaug(x;θ)=1−cos(h(A1(x)),h(A2(x))).
Appendix
Algorithm 1 Operational DSF update used for image optimization.
Model
Model type
Objective
Representation and score layers
Target models
ResNet
ConvNet classifier
classification
AvgPool/classifier features; conv. and residual-block layers
DETR
detection transformer
detection
encoder features; encoder projection layers
DINOv2 ViT
ViT encoder
self-supervised learning
final normalized features; late attention and MLP layers
External guidance model
CLIP
vision–language encoder
contrastive image–text learning
visual embedding; visual encoder layers
Appendix
Table 3: Architecture and objective coverage of DSF. The same DSF update is used across target discriminative models, while CLIP is used as an external guidance model.
Calibration source
8-bit scale
6-bit scale
4-bit scale
Real images
2.0
0.75
0.75
DSF
2.0
0.75
1.0
RegCache
1.5
1.0
1.25
Gaussian noise
2.0
1.25
1.5
MimiQ
2.0
1.25
0.75
Appendix
Table 4: Activation-scale multipliers used for zero-shot quantization diagnostics. Each value is selected from the calibration-source sweep used in Table 1 .
Figure 9: More images generated by ResNet50/ImageNet.
Figure 10: More images generated by DINOv2/LVD-124M.
We introduce LLaDA-Image, a unified framework that pairs a 6B Diffusion Transformer (DiT) trained from scratch with a frozen vision-language understanding module built on the LLaDA2.0-Mini diffusion language model backbone. Instead of relying heavily on paired image-text data from the beginning, we first build a strong visual generative prior through image-only pre-training and mid-training. The generation pipeline comprises 220M samples, 98 of which are real images. For efficient and scalable optimization, we use parameter-free RMSNorm throughout the DiT together with the Muon optimizer. The resulting unified model produces highly photorealistic images while accurately following fine-grained editing instructions. We further distill LLaDA-Image into LLaDA-Image-Turbo, enabling fast inference in 2-4 sampling steps. On Qwen-Image-Bench, LLaDA-Image achieves overall scores of 53.53 and 53.38 on the English and Chinese tracks, respectively, setting a new state-of-the-art among open-source models on both tracks. To support further research on capable and efficient generative models, we release our model weights, training code, and detailed recipes.
Diffusion models have achieved impressive results in generative tasks such as text-to-image synthesis, yet they often struggle to fully align outputs with nuanced user intent and maintain consistent aesthetic quality. Existing preference-based training methods such as Diffusion Direct Preference Optimization help address these issues, but obtain their supervision targets from the forward process q(xt−1∣xt,x0) derived from terminal samples, which is not directly aligned with the model's actual backward denoising transitions at each step. In this work, we introduce Direct Diffusion Score Preference Optimization (DDSPO), which defines stepwise preference supervision directly over backward denoising transitions through a contrastive policy pair, rather than relying on forward-process approximations from terminal samples. We propose two practical instantiations of the contrastive policy pair: training separate winning and losing models on preference data, and inducing a contrastive policy pair without additional training by using a pretrained reference model conditioned on an original prompt and a semantically degraded variant, requiring neither reward modeling nor manual annotations. Empirical results show that contrastive-policy-pair supervision is more effective than forward-process-based supervision across text-image alignment and aesthetic-quality tasks. Our implementation is available at: https://dohyun-as.github.io/DDSPO
Diffusion models have become a dominant paradigm for conditional image generation, yet existing approaches generally follow two directions: task-specific designs that can improve performance but limit generalization, and training-free loss guidance that compresses rich conditions into scalar objectives and applies stepwise guidance, leading to information bottlenecks and error accumulation along the sampling trajectory. Given the urgent need for an effective unified framework across diverse conditional image generation tasks, we propose Data Injection and Contrastive Trajectory Refinement (DICT), a training-free inference method that enhances conditional image generation without introducing task-dependent architectures. DICT introduces Data Injection, where noise-perturbed conditional signals are integrated into early denoising stages; by performing guided denoising on these injected signals, DICT adaptively selects and distills task-salient information from the raw condition, effectively preserving spatial richness and ensuring precise condition-to-generation alignment. Furthermore, DICT applies Contrastive Trajectory Refinement across adjacent denoising states, enabling pairwise comparisons that progressively improve sample quality. These designs keep inference simple while improving cross-task transfer under a unified diffusion formulation. Extensive experiments on conditional image generation tasks (e.g., style transfer, image super-resolution, and image deblurring) show consistent gains in fidelity and perceptual quality over representative task-specific and loss-guided baselines.