I-Max: Maximize the Resolution Potential of Pre-trained Rectified Flow Transformers with Projected Flow

Rectified Flow Transformers (RFTs) offer superior training and inference efficiency, making them likely the most viable direction for scaling up diffusion models. However, progress in generation resolution has been relatively slow due to data quality and training costs. Tuning-free resolution extrapolation presents an alternative, but current methods often reduce generative stability, limiting practical application. In this paper, we review existing resolution extrapolation methods and introduce the I-Max framework to maximize the resolution potential of Text-to-Image RFTs. I-Max features: (i) a novel Projected Flow strategy for stable extrapolation and (ii) an advanced inference toolkit for generalizing model knowledge to higher resolutions. Experiments with Lumina-Next-2K and Flux.1-dev demonstrate I-Max's ability to enhance stability in resolution extrapolation and show that it can bring image detail emergence and artifact correction, confirming the practical value of tuning-free resolution extrapolation.

Lumina-mGPT: Illuminate Flexible Photorealistic Text-to-Image Generation with Multimodal Generative Pretraining

We present Lumina-mGPT, a family of multimodal autoregressive models capable of various vision and language tasks, particularly excelling in generating flexible photorealistic images from text descriptions. Unlike existing autoregressive image generation approaches, Lumina-mGPT employs a pretrained decoder-only transformer as a unified framework for modeling multimodal token sequences. Our key insight is that a simple decoder-only transformer with multimodal Generative PreTraining (mGPT), utilizing the next-token prediction objective on massive interleaved text-image sequences, can learn broad and general multimodal capabilities, thereby illuminating photorealistic text-to-image generation. Building on these pretrained models, we propose Flexible Progressive Supervised Finetuning (FP-SFT) on high-quality image-text pairs to fully unlock their potential for high-aesthetic image synthesis at any resolution while maintaining their general multimodal capabilities. Furthermore, we introduce Ominiponent Supervised Finetuning (Omni-SFT), transforming Lumina-mGPT into a foundation model that seamlessly achieves omnipotent task unification. The resulting model demonstrates versatile multimodal capabilities, including visual generation tasks like flexible text-to-image generation and controllable generation, visual recognition tasks like segmentation and depth estimation, and vision-language tasks like multiturn visual question answering. Additionally, we analyze the differences and similarities between diffusion-based and autoregressive methods in a direct comparison.

VEnhancer: Generative Space-Time Enhancement for Video Generation

We present VEnhancer, a generative space-time enhancement framework that improves the existing text-to-video results by adding more details in spatial domain and synthetic detailed motion in temporal domain. Given a generated low-quality video, our approach can increase its spatial and temporal resolution simultaneously with arbitrary up-sampling space and time scales through a unified video diffusion model. Furthermore, VEnhancer effectively removes generated spatial artifacts and temporal flickering of generated videos. To achieve this, basing on a pretrained video diffusion model, we train a video ControlNet and inject it to the diffusion model as a condition on low frame-rate and low-resolution videos. To effectively train this video ControlNet, we design space-time data augmentation as well as video-aware conditioning. Benefiting from the above designs, VEnhancer yields to be stable during training and shares an elegant end-to-end training manner. Extensive experiments show that VEnhancer surpasses existing state-of-the-art video super-resolution and space-time super-resolution methods in enhancing AI-generated videos. Moreover, with VEnhancer, exisiting open-source state-of-the-art text-to-video method, VideoCrafter-2, reaches the top one in video generation benchmark -- VBench.

Triplet Knowledge Distillation

In Knowledge Distillation, the teacher is generally much larger than the student, making the solution of the teacher likely to be difficult for the student to learn. To ease the mimicking difficulty, we introduce a triplet knowledge distillation mechanism named TriKD. Besides teacher and student, TriKD employs a third role called anchor model. Before distillation begins, the pre-trained anchor model delimits a subspace within the full solution space of the target problem. Solutions within the subspace are expected to be easy targets that the student could mimic well. Distillation then begins in an online manner, and the teacher is only allowed to express solutions within the aforementioned subspace. Surprisingly, benefiting from accurate but easy-to-mimic hints, the student can finally perform well. After the student is well trained, it can be used as the new anchor for new students, forming a curriculum learning strategy. Our experiments on image classification and face recognition with various models clearly demonstrate the effectiveness of our method. Furthermore, the proposed TriKD is also effective in dealing with the overfitting issue. Moreover, our theoretical analysis supports the rationality of our triplet distillation.

Function-Consistent Feature Distillation

Feature distillation makes the student mimic the intermediate features of the teacher. Nearly all existing feature-distillation methods use L2 distance or its slight variants as the distance metric between teacher and student features. However, while L2 distance is isotropic w.r.t. all dimensions, the neural network's operation on different dimensions is usually anisotropic, i.e., perturbations with the same 2-norm but in different dimensions of intermediate features lead to changes in the final output with largely different magnitude. Considering this, we argue that the similarity between teacher and student features should not be measured merely based on their appearance (i.e., L2 distance), but should, more importantly, be measured by their difference in function, namely how later layers of the network will read, decode, and process them. Therefore, we propose Function-Consistent Feature Distillation (FCFD), which explicitly optimizes the functional similarity between teacher and student features. The core idea of FCFD is to make teacher and student features not only numerically similar, but more importantly produce similar outputs when fed to the later part of the same network. With FCFD, the student mimics the teacher more faithfully and learns more from the teacher. Extensive experiments on image classification and object detection demonstrate the superiority of FCFD to existing methods. Furthermore, we can combine FCFD with many existing methods to obtain even higher accuracy. Our codes are available at https://github.com/LiuDongyang6/FCFD.

Give Me Something to Eat: Referring Expression Comprehension with Commonsense Knowledge

Conventional referring expression comprehension (REF) assumes people to query something from an image by describing its visual appearance and spatial location, but in practice, we often ask for an object by describing its affordance or other non-visual attributes, especially when we do not have a precise target. For example, sometimes we say 'Give me something to eat'. In this case, we need to use commonsense knowledge to identify the objects in the image. Unfortunately, these is no existing referring expression dataset reflecting this requirement, not to mention a model to tackle this challenge. In this paper, we collect a new referring expression dataset, called KB-Ref, containing 43k expressions on 16k images. In KB-Ref, to answer each expression (detect the target object referred by the expression), at least one piece of commonsense knowledge must be required. We then test state-of-the-art (SoTA) REF models on KB-Ref, finding that all of them present a large drop compared to their outstanding performance on general REF datasets. We also present an expression conditioned image and fact attention (ECIFA) network that extract information from correlated image regions and commonsense knowledge facts. Our method leads to a significant improvement over SoTA REF models, although there is still a gap between this strong baseline and human performance. The dataset and baseline models will be released.