Jianfei Song

Wenshuo Chen, Kuimou Yu, Bowen Tian et al.

Modern text-to-image models have achieved strong visual synthesis, yet remain unreliable when prompts require implicit visual constraints, relational reasoning, or external knowledge. Existing retrieval-augmented and agentic generation methods mitigate this issue by acquiring external knowledge, references, or refined prompts for the current request, yet they typically treat each generation as an isolated episode and do not systematically preserve past successes or failures for future use. In this work, we ask whether a text-to-image system can continually improve from its own generation experience without updating the underlying generator. We propose MemoGen, a training-free framework that augments existing image generators with an agentic evolution layer. For each task, MemoGen explicitly infers visual requirements, retrieves external evidence and references when necessary, translates them into executable generation constraints, evaluates the generated result, and stores task understanding, reference choices, visual feedback, successful strategies, and failure lessons as reusable experience memory. Across evolution rounds, the agent retrieves relevant experience to improve similar future generations, selectively repairing previously failed cases while preserving successful ones, thereby enabling test-time self-evolution without parameter updates. Extensive experiments on knowledge-intensive and reasoning-oriented benchmarks demonstrate the effectiveness of this paradigm: after only two evolution rounds, MemoGen built upon the open-source Qwen-Image backbone surpasses strong proprietary systems such as Nano Banana Pro and GPT-Image-1 on WISE and Mind-Bench, showing that explicit experience memory can serve as a powerful continual learning signal for reliable text-to-image generation.

Chunyu Xie, Heng Cai, Jincheng Li et al.

Vision-language pre-training (VLP) on large-scale datasets has shown premier performance on various downstream tasks. In contrast to plenty of available benchmarks with English corpus, large-scale pre-training datasets and downstream datasets with Chinese corpus remain largely unexplored. In this work, we build a large-scale high-quality Chinese Cross-Modal Benchmark named CCMB for the research community, which contains the currently largest public pre-training dataset Zero and five human-annotated fine-tuning datasets for downstream tasks. Zero contains 250 million images paired with 750 million text descriptions, plus two of the five fine-tuning datasets are also currently the largest ones for Chinese cross-modal downstream tasks. Along with the CCMB, we also develop a VLP framework named R2D2, applying a pre-Ranking + Ranking strategy to learn powerful vision-language representations and a two-way distillation method (i.e., target-guided Distillation and feature-guided Distillation) to further enhance the learning capability. With the Zero and the R2D2 VLP framework, we achieve state-of-the-art performance on twelve downstream datasets from five broad categories of tasks including image-text retrieval, image-text matching, image caption, text-to-image generation, and zero-shot image classification. The datasets, models, and codes are available at https://github.com/yuxie11/R2D2

Lin Yao, Jianfei Song, Ruizhuo Xu et al.

Historically lower-level tasks such as automatic speech recognition (ASR) and speaker identification are the main focus in the speech field. Interest has been growing in higher-level spoken language understanding (SLU) tasks recently, like sentiment analysis (SA). However, improving performances on SLU tasks remains a big challenge. Basically, there are two main methods for SLU tasks: (1) Two-stage method, which uses a speech model to transfer speech to text, then uses a language model to get the results of downstream tasks; (2) One-stage method, which just fine-tunes a pre-trained speech model to fit in the downstream tasks. The first method loses emotional cues such as intonation, and causes recognition errors during ASR process, and the second one lacks necessary language knowledge. In this paper, we propose the Wave BERT (WaBERT), a novel end-to-end model combining the speech model and the language model for SLU tasks. WaBERT is based on the pre-trained speech and language model, hence training from scratch is not needed. We also set most parameters of WaBERT frozen during training. By introducing WaBERT, audio-specific information and language knowledge are integrated in the short-time and low-resource training process to improve results on the dev dataset of SLUE SA tasks by 1.15% of recall score and 0.82% of F1 score. Additionally, we modify the serial Continuous Integrate-and-Fire (CIF) mechanism to achieve the monotonic alignment between the speech and text modalities.

Haozhe Jia, Jianfei Song, Yuan Zhang et al.

We present ECHO, an edge--cloud framework for language-driven whole-body control of humanoid robots. A cloud-hosted diffusion-based text-to-motion generator synthesizes motion references from natural language instructions, while an edge-deployed reinforcement-learning tracker executes them in closed loop on the robot. The two modules are bridged by a compact, robot-native 38-dimensional motion representation that encodes joint angles, root planar velocity, root height, and a continuous 6D root orientation per frame, eliminating inference-time retargeting from human body models and remaining directly compatible with low-level PD control. The generator adopts a 1D convolutional UNet with cross-attention conditioned on CLIP-encoded text features; at inference, DDIM sampling with 10 denoising steps and classifier-free guidance produces motion sequences in approximately one second on a cloud GPU. The tracker follows a Teacher--Student paradigm: a privileged teacher policy is distilled into a lightweight student equipped with an evidential adaptation module for sim-to-real transfer, further strengthened by morphological symmetry constraints and domain randomization. An autonomous fall recovery mechanism detects falls via onboard IMU readings and retrieves recovery trajectories from a pre-built motion library. We evaluate ECHO on a retargeted HumanML3D benchmark, where it achieves strong generation quality (FID 0.029, R-Precision Top-1 0.686) under a unified robot-domain evaluator, while maintaining high motion safety and trajectory consistency. Real-world experiments on a Unitree G1 humanoid demonstrate stable execution of diverse text commands with zero hardware fine-tuning.

Haozhe Jia, Honglei Jin, Yuan Zhang et al.

Natural language is an intuitive interface for humanoid robots, yet streaming whole-body control requires control representations that are executable now and anticipatory of future physical transitions. Existing language-conditioned humanoid systems typically generate kinematic references that a low-level tracker must repair reactively, or use latent/action policies whose outputs do not explicitly encode upcoming contact changes, support transfers, and balance preparation. We propose \textbf{DAJI} (\emph{Dynamics-Aligned Joint Intent}), a hierarchical framework that learns an anticipatory joint-intent interface between language generation and closed-loop control. DAJI-Act distills a future-aware teacher into a deployable diffusion action policy through student-driven rollouts, while DAJI-Flow autoregressively generates future intent chunks from language and intent history. Experiments show that DAJI achieves strong results in anticipatory latent learning, single-instruction generation, and streaming instruction following, reaching 94.42\% rollout success on HumanML3D-style generation and 0.152 subsequence FID on BABEL.

Lijun Liu, Jiali Yang, Jianfei Song et al.

Recombinant adeno-associated virus (rAAV) vectors have revolutionized gene therapy, but their broad tropism and suboptimal transduction efficiency limit their clinical applications. To overcome these limitations, researchers have focused on designing and screening capsid libraries to identify improved vectors. However, the large sequence space and limited resources present challenges in identifying viable capsid variants. In this study, we propose an end-to-end diffusion model to generate capsid sequences with enhanced viability. Using publicly available AAV2 data, we generated 38,000 diverse AAV2 viral protein (VP) sequences, and evaluated 8,000 for viral selection. The results attested the superiority of our model compared to traditional methods. Additionally, in the absence of AAV9 capsid data, apart from one wild-type sequence, we used the same model to directly generate a number of viable sequences with up to 9 mutations. we transferred the remaining 30,000 samples to the AAV9 domain. Furthermore, we conducted mutagenesis on AAV9 VP hypervariable regions VI and V, contributing to the continuous improvement of the AAV9 VP sequence. This research represents a significant advancement in the design and functional validation of rAAV vectors, offering innovative solutions to enhance specificity and transduction efficiency in gene therapy applications.

Jiahong Wu, Jianfei Lu, Xinxin Kang et al.

Panoramic segmentation is a scene where image segmentation tasks is more difficult. With the development of CNN networks, panoramic segmentation tasks have been sufficiently developed.However, the current panoramic segmentation algorithms are more concerned with context semantics, but the details of image are not processed enough. Moreover, they cannot solve the problems which contains the accuracy of occluded object segmentation,little object segmentation,boundary pixel in object segmentation etc. Aiming to address these issues, this paper presents some useful tricks. (a) By changing the basic segmentation model, the model can take into account the large objects and the boundary pixel classification of image details. (b) Modify the loss function so that it can take into account the boundary pixels of multiple objects in the image. (c) Use a semi-supervised approach to regain control of the training process. (d) Using multi-scale training and reasoning. All these operations named AinnoSeg, AinnoSeg can achieve state-of-art performance on the well-known dataset ADE20K.

Xinyu Fan, Faen Zhang, Jianfei Song et al.

A fine-grained provenance-based access control policy model is proposed in this paper, in order to improve the express performance of existing model. This method employs provenance as conditions to determine whether a piece of data can be accessed because historical operations performed on data could reveal clues about its sensitivity and vulnerability. Particularly, our proposed work provides a four-valued decision set which allows showing status to match a restriction particularly. This framework consists of target policy, access control policy, and policy algebras. With the complete definition and algebra system construction, a practical fine-grained access control policy model is developed.

Xinyu Fan, Faen Zhang, Jianfei Song et al.

Even though the idea of partitioning provenance graphs for access control was previously proposed, employing segments of the provenance DAG for fine-grained access control to provenance data has not been thoroughly explored. Hence, we take segments of a provenance graph, based on the extended OPM, and defined use a variant of regular expressions, and utilize them in our fine-grained access control language. It can not only return partial graphs to answer access requests but also introduce segments as restrictions in order to screen targeted data.

Faen Zhang, Xinyu Fan, Guo Ai et al.

Face detection has witnessed significant progress due to the advances of deep convolutional neural networks (CNNs). Its central issue in recent years is how to improve the detection performance of tiny faces. To this end, many recent works propose some specific strategies, redesign the architecture and introduce new loss functions for tiny object detection. In this report, we start from the popular one-stage RetinaNet approach and apply some recent tricks to obtain a high performance face detector. Specifically, we apply the Intersection over Union (IoU) loss function for regression, employ the two-step classification and regression for detection, revisit the data augmentation based on data-anchor-sampling for training, utilize the max-out operation for classification and use the multi-scale testing strategy for inference. As a consequence, the proposed face detection method achieves state-of-the-art performance on the most popular and challenging face detection benchmark WIDER FACE dataset.