Oracle Bone Script (OBS) is a vital treasure of human civilization, rich in insights from ancient societies. However, the evolution of written language over millennia complicates its decipherment. In this paper, we propose V-Oracle, an innovative framework that utilizes Large Multi-modal Models (LMMs) for interpreting OBS. V-Oracle applies principles of pictographic character formation and frames the task as a visual question-answering (VQA) problem, establishing a multi-step reasoning chain. It proposes a multi-dimensional data augmentation for synthesizing high-quality OBS samples, and also implements a multi-phase oracle alignment tuning to improve LMMs’ visual reasoning capabilities. Moreover, to bridge the evaluation gap in the OBS field, we further introduce Oracle-Bench, a comprehensive benchmark that emphasizes process-oriented assessment and incorporates both standard and out-of-distribution setups for realistic evaluation. Extensive experimental results can demonstrate the effectiveness of our method in providing quantitative analyses and superior deciphering capability.

Runqi Qiao, Qiuna Tan, Guanting Dong, MinhuiWu MinhuiWu, Chong Sun, Xiaoshuai Song, Jiapeng Wang, Zhuoma GongQue, Shanglin Lei, YiFan Zhang, Zhe Wei, Miaoxuan Zhang, Runfeng Qiao, Xiao Zong, Yida Xu, Peiqing Yang, Zhimin Bao, Muxi Diao, Chen Li, Honggang Zhang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Existing multimodal generative models fall short as qualified design copilots, as they often struggle to generate imaginative outputs once instructions are less detailed or lack the ability to maintain consistency with the provided references. In this work, we introduce WeGen, a model that unifies multimodal generation and understanding, and promotes their interplay in iterative generation. It can generate diverse results with high creativity for less detailed instructions. And it can progressively refine prior generation results or integrating specific contents from references following the instructions in its chat with users. During this process, it is capable of preserving consistency in the parts that the user is already satisfied with. To this end, we curate a large-scale dataset, extracted from Internet videos, containing rich object dynamics and auto-labeled dynamics descriptions by advanced foundation models to date. These two information are interleaved into a single sequence to enable WeGen to learn consistency-aware generation where the specified dynamics are generated while the consistency of unspecified content is preserved aligned with instructions. Besides, we introduce a prompt self-rewriting mechanism to enhance generation diversity. Extensive experiments demonstrate the effectiveness of unifying multimodal understanding and generation in WeGen and show it achieves state-of-the-art performance across various visual generation benchmarks. These also demonstrate the potential of WeGen as a user-friendly design copilot as desired. The code and models will be available at https://github.com/hzphzp/WeGen.

The rapid advancement of face forgery techniques has introduced a growing variety of forgeries. Incremental Face Forgery Detection (IFFD), involving gradually adding new forgery data to fine-tune the previously trained model, has been introduced as a promising strategy to deal with evolving forgery methods. However, a naively trained IFFD model is prone to catastrophic forgetting when new forgeries are integrated, as treating all forgeries as a single “Fake” class in the Real/Fake classification can cause different forgery types overriding one another, thereby resulting in the forgetting of unique characteristics from earlier tasks and limiting the model’s effectiveness in learning forgery specificity and generality. In this paper, we propose to stack the latent feature distributions of previous and new tasks brick by brick, i.e., achieving aligned feature isolation. In this manner, we aim to preserve learned forgery information and accumulate new knowledge by minimizing distribution overriding, thereby mitigating catastrophic forgetting. To achieve this, we first introduce Sparse Uniform Replay (SUR) to obtain the representative subsets that could be treated as the uniformly sparse versions of the previous global distributions. We then propose a Latent-space Incremental Detector (LID) that leverages SUR data to isolate and align distributions. For evaluation, we construct a more advanced and comprehensive benchmark tailored for IFFD. The leading experimental results validate the superiority of our method. Code is available at https://github.com/beautyremain/SUR-LID .

Recent advancements in multimodal vision models have highlighted limitations in late-stage feature fusion and suboptimal query selection for hybrid prompts open-world segmentation, alongside constraints from caption-derived vocabularies. To address these challenges, we propose Prompt-DINO, a text-guided visual Prompt DINO framework featuring three key innovations. First, we introduce an early fusion mechanism that unifies text/visual prompts and backbone features at the initial encoding stage, enabling deeper cross-modal interactions to resolve semantic ambiguities. Second, we design order-aligned query selection for DETR-based architectures, explicitly optimizing the structural alignment between text and visual queries during decoding to enhance semantic-spatial consistency. Third, we develop a generative data engine powered by the Recognize Anything via Prompting (RAP) model, which synthesizes 0.5B diverse training instances through a dual-path cross-verification pipeline, reducing label noise by 80.5% compared to conventional approaches. Extensive experiments demonstrate that Prompt-DINO achieves state-of-the-art performance on open-world detection benchmarks while significantly expanding semantic coverage beyond fixed-vocabulary constraints. Our work establishes a new paradigm for scalable multimodal detection and data generation in open-world scenarios. Data&Code are available at https://github.com/WeChatCV/WeVisionOne.

The diffusion models (DMs) have demonstrated the remarkable capability of generating images via learning the noised score function of data distribution. Current DM sampling techniques typically rely on first-order Langevin dynamics at each noise level, with efforts concentrated on refining inter-level denoising strategies. While leveraging additional second-order Hessian geometry to enhance the sampling quality of Langevin is a common practice in Markov chain Monte Carlo (MCMC), the naive attempts to utilize Hessian geometry in high-dimensional DMs lead to quadratic-complexity computational costs, rendering them non-scalable. In this work, we introduce a novel Levenberg-Marquardt-Langevin (LML) method that approximates the diffusion Hessian geometry in a training-free manner, drawing inspiration from the celebrated Levenberg-Marquardt optimization algorithm. Our approach introduces two key innovations: (1) A low-rank approximation of the diffusion Hessian, leveraging the DMs' inherent structure and circumventing explicit quadratic-complexity computations; (2) A damping mechanism to stabilize the approximated Hessian. This LML approximated Hessian geometry enables the diffusion sampling to execute more accurate steps and improve the image generation quality. We further conduct a theoretical analysis to substantiate the approximation error bound of low-rank approximation and the convergence property of the damping mechanism. Extensive experiments across multiple pretrained DMs validate that the LML method significantly improves image generation quality, with negligible computational overhead.

Recent Diffusion models (DMs) advancements have explored incorporating the second-order diffusion Fisher information (DF), defined as the negative Hessian of log density, into various downstream tasks and theoretical analysis. However, current practices typically approximate the diffusion Fisher by applying auto-differentiation to the learned score network. This black-box method, though straightforward, lacks any accuracy guarantee and is time-consuming. In this paper, we show that the diffusion Fisher actually resides within a space spanned by the outer products of score and initial data. Based on the outer-product structure, we develop two efficient approximation algorithms to access the trace and matrix-vector multiplication of DF, respectively. These algorithms bypass the auto-differentiation operations with time-efficient vector-product calculations. Furthermore, we establish the approximation error bounds for the proposed algorithms. Experiments in likelihood evaluation and adjoint optimization demonstrate the superior accuracy and reduced computational cost of our proposed algorithms. Additionally, based on the novel outer-product formulation of DF, we design the first numerical verification experiment for the optimal transport property of the general PF-ODE deduced map.

Diffusion models have driven the advancement of vision generation over the past years. However, it is often difficult to apply these large models in downstream tasks, due to massive fine-tuning cost. Recently, Low-Rank Adaptation (LoRA) has been applied for efficient tuning of diffusion models. Unfortunately, the capabilities of LoRA-tuned diffusion models are limited, since the same LoRA is used for different timesteps of the diffusion process. To tackle this problem, we introduce a general and concise TimeStep Master (TSM) paradigm with two key fine-tuning stages. In the fostering stage (1-stage), we apply different LoRAs to fine-tune the diffusion model at different timestep intervals. This results in different TimeStep LoRA experts that can effectively capture different noise levels. In the assembling stage (2-stage), we design a novel asymmetrical mixture of TimeStep LoRA experts, via core-context collaboration of experts at multi-scale intervals. For each timestep, we leverage TimeStep LoRA expert within the smallest interval as the core expert without gating, and use experts within the bigger intervals as the context experts with time-dependent gating. Consequently, our TSM can effectively model the noise level via the expert in the finest interval, and adaptively integrate contexts from the experts of other scales, boosting the versatility of diffusion models. To show the effectiveness of our TSM paradigm, we conduct extensive experiments on three typical and popular LoRA-related tasks of diffusion models, including domain adaptation, post-pretraining, and model distillation. Our TSM achieves the state-of-the-art results on all these tasks, throughout various model structures (UNet, DiT and MM-DiT) and visual data modalities (Image, Video), showing its remarkable generalization capacity.

Recent advancements in Large Multimodal Models have demonstrated impressive performance in various tasks. However, their capabilities in error detection and resolution for Optical Character Recognition (OCR) remain underexplored. To address this gap, we construct the first visual instruction tuning dataset specifically for detailed OCR error analysis. Building on this foundation, we develop a universal, plug-and-play OCR-Critic model that incorporates three novel dynamic alignment strategies. These strategies systematically mitigate LMMs' weaknesses in OCR tasks by providing coarse-to-fine error feedback. To comprehensively evaluate these capabilities, we introduce OCR-ERROR, a benchmark designed to assess LMMs' ability to detect and categorize OCR errors, covering two task types, diverse error categories, and 2,400 rigorously validated samples. Experimental results show that OCR-Critic effectively identifies fine-grained OCR errors across multiple domains. With the integration of our dynamic alignment strategies, the LMM further achieves substantial performance gains on four prominent benchmarks, demonstrating both versatility and effectiveness.

Learning intrinsic bias from limited data has been considered the main reason for the failure of deepfake detection with generalizability. Apart from the discovered content and specific-forgery bias, we reveal a novel spatial bias, where detectors inertly anticipate observing structural forgery clues appearing at the image center, also can lead to the poor generalization of existing methods. We present ED4, a simple and effective strategy, to address aforementioned biases explicitly at the data level in a unified framework rather than implicit disentanglement via network design. In particular, we develop ClockMix to produce facial structure preserved mixtures with arbitrary samples, which allows the detector to learn from an exponentially extended data distribution with much more diverse identities, backgrounds, local manipulation traces, and the co-occurrence of multiple forgery artifacts. We further propose the Adversarial Spatial Consistency Module (AdvSCM) to prevent extracting features with spatial bias, which adversarially generates spatial-inconsistent images and constrains their extracted feature to be consistent. As a model-agnostic debiasing strategy, ED4 is plug-and-play: it can be integrated with various deepfake detectors to obtain significant benefits. We conduct extensive experiments to demonstrate its effectiveness and superiority over existing deepfake detection approaches. Code is available at https://github.com/beautyremain/ED4.

Most diffusion models assume that the reverse process adheres to a Gaussian distribution. However, this approxi-mation has not been rigorously validated, especially at sin-gularities, where t = 0 and t = 1. Improperly dealing with such singularities leads to an average brightness is-sue in applications, and limits the generation of images with extreme brightness or darkness. We primarily focus on tackling singularities from both theoretical and practi-cal perspectives. Initially, we establish the error bounds for the reverse process approximation, and showcase its Gaussian characteristics at singularity time steps. Based on this theoretical insight, we confirm the singularity at t = 1 is conditionally removable while it at t = 0 is an inherent property. Upon these significant conclusions, we propose a novel plug-and-play method SingDiffusion to address the initial singular time step sampling, which not only effectively resolves the average brightness issue for a wide range of diffusion models without extra training efforts, but also enhances their generation capability in achieving notable lower FID scores.

The generalization ability of deepfake detectors is vital for their applications in real-world scenarios. One effective solution to enhance this ability is to train the models with manually-blended data, which we termed ''blendfake'', encouraging models to learn generic forgery artifacts like blending boundary. Interestingly, current SoTA methods utilize blendfake $\textit{without}$ incorporating any deepfake data in their training process. This is likely because previous empirical observations suggest that vanilla hybrid training (VHT), which combines deepfake and blendfake data, results in inferior performance to methods using only blendfake data (so-called "1+1<2"). Therefore, a critical question arises: Can we leave deepfake behind and rely solely on blendfake data to train an effective deepfake detector? Intuitively, as deepfakes also contain additional informative forgery clues ($\textit{e.g.,}$ deep generative artifacts), excluding all deepfake data in training deepfake detectors seems counter-intuitive. In this paper, we rethink the role of blendfake in detecting deepfakes and formulate the process from "real to blendfake to deepfake" to be a $\textit{progressive transition}$. Specifically, blendfake and deepfake can be explicitly delineated as the oriented pivot anchors between "real-to-fake" transitions. The accumulation of forgery information should be oriented and progressively increasing during this transition process. To this end, we propose an $\underline{O}$riented $\underline{P}$rogressive $\underline{R}$egularizor (OPR) to establish the constraints that compel the distribution of anchors to be discretely arranged. Furthermore, we introduce feature bridging to facilitate the smooth transition between adjacent anchors. Extensive experiments confirm that our design allows leveraging forgery information from both blendfake and deepfake effectively and comprehensively. Code is available at https://github.com/beautyremain/ProDet.

The inversion of diffusion model sampling, which aims to find the corresponding initial noise of a sample, plays a critical role in various tasks. Recently, several heuristic exact inversion samplers have been proposed to address the inexact inversion issue in a training-free manner. However, the theoretical properties of these heuristic samplers remain unknown and they often exhibit mediocre sampling quality. In this paper, we introduce a generic formulation, \emph{Bidirectional Explicit Linear Multi-step} (BELM) samplers, of the exact inversion samplers, which includes all previously proposed heuristic exact inversion samplers as special cases. The BELM formulation is derived from the variable-stepsize-variable-formula linear multi-step method via integrating a bidirectional explicit constraint. We highlight this bidirectional explicit constraint is the key of mathematically exact inversion. We systematically investigate the Local Truncation Error (LTE) within the BELM framework and show that the existing heuristic designs of exact inversion samplers yield sub-optimal LTE. Consequently, we propose the Optimal BELM (O-BELM) sampler through the LTE minimization approach. We conduct additional analysis to substantiate the theoretical stability and global convergence property of the proposed optimal sampler. Comprehensive experiments demonstrate our O-BELM sampler establishes the exact inversion property while achieving high-quality sampling. Additional experiments in image editing and image interpolation highlight the extensive potential of applying O-BELM in varying applications.

Weak-shot Object Detection methods exploit a fully-annotated source dataset to facilitate the detection performance on the target dataset which only contains image-level labels for novel categories. To bridge the gap between these two datasets, we aim to transfer the object knowledge between the source (S) and target (T) datasets in a bi-directional manner. We propose a novel Knowledge Transfer (KT) loss which simultaneously distills the knowledge of objectness and class entropy from a proposal generator trained on the S dataset to optimize a multiple instance learning module on the T dataset. By jointly optimizing the classification loss and the proposed KT loss, the multiple instance learning module effectively learns to classify object proposals into novel categories in the T dataset with the transferred knowledge from base categories in the S dataset. Noticing the predicted boxes on the T dataset can be regarded as an extension for the original annotations on the S dataset to refine the proposal generator in return, we further propose a novel Consistency Filtering (CF) method to reliably remove inaccurate pseudo labels by evaluating the stability of the multiple instance learning module upon noise injections. Via mutually transferring knowledge between the S and T datasets in an iterative manner, the detection performance on the target dataset is significantly improved. Extensive experiments on public benchmarks validate that the proposed method performs favourably against the state-of-the-art methods without increasing the model parameters or inference computational complexity.

Recent progress on neural architecture search (NAS) has demonstrated exciting results on automating deep network architecture designs. In order to overcome the unaffordable complexity of training each candidate architecture from scratch, the state-of-the-art one-shot NAS approaches adopt a weight-sharing strategy to improve training efficiency. Although the computational cost is greatly reduced, such oneshot process introduces a severe weight coupling problem that largely degrades the evaluation accuracy of each candidate. The existing approaches often address the problem by shrinking the search space, model distillation, or fewshot training. Instead, in this paper, we propose a novel distribution consistent one-shot neural architecture search algorithm. We first theoretically investigate how the weight coupling problem affects the network searching performance from a parameter distribution perspective, and then propose a novel supernet training strategy with a Distribution Consistent Constraint that can provide a good measurement for the extent to which two architectures can share weights. Our strategy optimizes the supernet through iteratively inferring network weights and corresponding local sharing states. Such joint optimization of supernet's weights and topologies can diminish the discrepancy between the weights inherited from the supernet and the ones that are trained with a stand-alone model. As a result, it enables a more accurate model evaluation phase and leads to a better searching performance. We conduct extensive experiments on benchmark datasets with multiple searching spaces. The resulting architecture achieves superior performance over the current state-of-the-art NAS algorithms with comparable search costs, which demonstrates the efficacy of our approach.

We propose HyFRIS-Net to jointly estimate the hybrid reflectance and illumination models, as well as the refined face shape from a single unconstrained face image in a pre-defined texture space. The proposed hybrid reflectance and illumination representation ensure photometric face appearance modeling in both parametric and non-parametric spaces for efficient learning. While forcing the reflectance consistency constraint for the same person and face identity constraint for different persons, our approach recovers an occlusion-free face albedo with disambiguated color from the illumination color. Our network is trained in a self-evolving manner to achieve general applicability on real-world data. We conduct comprehensive qualitative and quantitative evaluations with state-of-the-art methods to demonstrate the advantages of HyFRIS-Net in modeling photo-realistic face albedo, illumination, and shape.

Human pose transfer has been becoming one of the emerging research topics in recent years. However, state-of-the-art results are still far from satisfactory. One main reason is that these end-to-end methods are often blindly trained without the semantic understanding of its content. In this paper, we propose a novel method for human pose transfer with consideration of the semantic part-based representation of a human. In particular, we propose to segment the human body into multiple parts, and each of them represents a semantic region of a human. With the proposed part-based layer generators, a high-quality result is guaranteed for each local semantic region. We design a three-stage hierarchical framework to fuse local representations into the final result in a coarse-to-fine manner, which provides adaptive attention for global consistency and local details, respectively. Via exploiting spatial guidance from 3D human model through the framework, our method can naturally handle the ambiguity of self-occlusions which always causes artifacts in previous methods. With semantic-aware and spatial-aware representations, our method outperforms previous approaches quantitatively and qualitatively in better handling self-occlusions, fine detail preservation/synthesis and a higher resolution result.

We propose to address the face reconstruction in the wild by using a multi-metric regression network, MMFace, to align a 3D face morphable model (3DMM) to an input image. The key idea is to utilize a volumetric sub-network to estimate an intermediate geometry representation, and a parametric sub-network to regress the 3DMM parameters. Our parametric sub-network consists of identity loss, expression loss, and pose loss which greatly improves the aligned geometry details by incorporating high level loss functions directly defined in the 3DMM parametric spaces. Our high-quality reconstruction is robust under large variations of expressions, poses, illumination conditions, and even with large partial occlusions. We evaluate our method by comparing the performance with state-of-the-art approaches on latest 3D face dataset LS3D-W and Florence. We achieve significant improvements both quantitatively and qualitatively. Due to our high-quality reconstruction, our method can be easily extended to generate high-quality geometry sequences for video inputs.

We present a physically-based approach for simulating makeup in face images. The key idea is to decompose the face image into intrinsic image layers - namely albedo, diffuse shading, and specular highlights - which are each differently affected by cosmetics, and then manipulate each layer according to corresponding models of reflectance. Accurate intrinsic image decompositions for faces are obtained with the help of human face priors, including statistics on skin reflectance and facial geometry. The intrinsic image layers are then transformed in appearance according to measured optical properties of cosmetics and proposed adaptations of physically-based reflectance models. With this approach, realistic results are generated in a manner that preserves the personal appearance features and lighting conditions of the target face while not requiring detailed geometric and reflectance measurements. We demonstrate this technique on various forms of cosmetics including foundation, blush, lipstick, and eye shadow. Results on both images and videos exhibit a close approximation to ground truth and compare favorably to existing techniques.

We present a deep learning-based method for removing makeup effects (de-makeup) in a face image. This problem poses a major challenge due to obscuring of the underlying facial features by cosmetics, which is very important in multimedia applications in the field of security, entertainment, and social networking. To address this task, we propose the bidirectional tunable de-makeup network (BTD-Net), which jointly learns the makeup process to aid in learning the de-makeup process. For tractable learning of the makeup process, which is a one-to-many mapping determined by the cosmetics that are applied, we introduce a latent variable that reflects the makeup style. This latent variable is extracted in the de-makeup process and used as a condition on the makeup process to constrain the one-to-many mapping to a specific solution. Through extensive experiments, our proposed BTD-Net is found to surpass the state-of-art techniques in estimating realistic non-makeup faces that correspond to the input makeup images. We additionally show that applications such as tuning the amount of makeup can be enhanced through the use of this method.

We present a method for radiometric calibration of cameras from a single image that contains a human face. This technique takes advantage of a low-rank property that exists among certain skin albedo gradients because of the pigments within the skin. This property becomes distorted in images that are captured with a non-linear camera response function, and we perform radiometric calibration by solving for the inverse response function that best restores this low-rank property in an image. Although this work makes use of the color properties of skin pigments, we show that this calibration is unaffected by the color of scene illumination or the sensitivities of the cameras color filters. Our experiments validate this approach on a variety of images containing human faces, and show that faces can provide an important source of calibration data in images where existing radiometric calibration techniques perform poorly.

We present a method for removing specular highlight reflections in facial images that may contain varying illumination colors. This is accurately achieved through the use of physical and statistical properties of human skin and faces. We employ a melanin and hemoglobin based model to represent the diffuse color variations in facial skin, and utilize this model to constrain the highlight removal solution in a manner that is effective even for partially saturated pixels. The removal of highlights is further facilitated through estimation of directionally variant illumination colors over the face, which is done while taking advantage of a statistically-based approximation of facial geometry. An important practical feature of the proposed method is that the skin color model is utilized in a way that does not require color calibration of the camera. Moreover, this approach does not require assumptions commonly needed in previous highlight removal techniques, such as uniform illumination color or piecewise-constant surface colors. We validate this technique through comparisons to existing methods for removing specular highlights.

We present a method that enhances the performance of depth-from-defocus (DFD) through the use of shading information. DFD suffers from important limitations--namely coarse shape reconstruction and poor accuracy on textureless surfaces--that can be overcome with the help of shading. We integrate both forms of data within a Bayesian framework that capitalizes on their relative strengths. Shading data, however, is challenging to accurately recover from surfaces that contain texture. To address this issue, we propose an iterative technique that utilizes depth information to improve shading estimation, which in turn is used to elevate depth estimation in the presence of textures. The shading estimation can be performed in general scenes with unknown illumination using an approximate estimate of scene lighting. With this approach, we demonstrate improvements over existing DFD techniques, as well as effective shape reconstruction of textureless surfaces.

We present a method for symmetric stereo matching in which outliers from occlusions, texture-less regions, and repeated patterns are handled in a soft and adaptive manner. Rather than making binary outlier decisions, our model incorporates continuous-valued confidence weights that account for outlier likelihood, to promote robustness in disparity estimation. In contrast to previous outlier labeling techniques that fix the labels at the start of optimization, our method iteratively updates our outlier confidence weights as the matching results are gradually refined. By doing this, errors in an initial labeling can be rectified in the matching process. Our model is optimized in an Expectation-Maximization framework that efficiently produces continuous disparity estimates. This approach provides a good combination of accuracy and speed. Experiments show that our method compares favorably to prior outlier labeling techniques on the Middlebury benchmark, and that it can generate high-quality reconstruction for outdoor images with much more complex occlusions.

We present a method for simulating makeup in a face image. To generate realistic results without detailed geometric and reflectance measurements of the user, we propose to separate the image into intrinsic image layers and alter them according to proposed adaptations of physically-based reflectance models. Through this layer manipulation, the measured properties of cosmetic products are applied while preserving the appearance characteristics and lighting conditions of the target face. This approach is demonstrated on various forms of cosmetics including foundation, blush, lipstick, and eye shadow. Experimental results exhibit a close approximation to ground truth images, without artifacts such as transferred personal features and lighting effects that degrade the results of image-based makeup transfer methods.

We present a method that enhances the performance of depth-from-defocus (DFD) through the use of shading information. DFD suffers from important limitations - namely coarse shape reconstruction and poor accuracy on texture less surfaces - that can be overcome with the help of shading. We integrate both forms of data within a Bayesian framework that capitalizes on their relative strengths. Shading data, however, is challenging to recover accurately from surfaces that contain texture. To address this issue, we propose an iterative technique that utilizes depth information to improve shading estimation, which in turn is used to elevate depth estimation in the presence of textures. With this approach, we demonstrate improvements over existing DFD techniques, as well as effective shape reconstruction of texture less surfaces.