Yuhao Luo

Yansong Qu, Zihao Sheng, Zilin Huang et al.

Reinforcement Learning (RL) has emerged as a dominant paradigm for end-to-end autonomous driving (AD). However, RL suffers from sample inefficiency and a lack of semantic interpretability in complex scenarios. Foundation Models, particularly Vision-Language Models (VLMs), can mitigate this by offering rich, context-aware knowledge, yet their high inference latency hinders deployment in high-frequency RL training loops. To bridge this gap, we present Found-RL, a platform tailored to efficiently enhance RL for AD using foundation models. A core innovation is the asynchronous batch inference framework, which decouples heavy VLM reasoning from the simulation loop, effectively resolving latency bottlenecks to support real-time learning. We introduce diverse supervision mechanisms: Value-Margin Regularization (VMR) and Advantage-Weighted Action Guidance (AWAG) to effectively distill expert-like VLM action suggestions into the RL policy. Additionally, we adopt high-throughput CLIP for dense reward shaping. We address CLIP's dynamic blindness via Conditional Contrastive Action Alignment, which conditions prompts on discretized speed/command and yields a normalized, margin-based bonus from context-specific action-anchor scoring. Found-RL provides an end-to-end pipeline for fine-tuned VLM integration and shows that a lightweight RL model can achieve near-VLM performance compared with billion-parameter VLMs while sustaining real-time inference (approx. 500 FPS). Code, data, and models will be publicly available at https://github.com/ys-qu/found-rl.

Zhiyuan Yan, Yuhao Luo, Siwei Lyu et al.

Deepfake detection faces a critical generalization hurdle, with performance deteriorating when there is a mismatch between the distributions of training and testing data. A broadly received explanation is the tendency of these detectors to be overfitted to forgery-specific artifacts, rather than learning features that are widely applicable across various forgeries. To address this issue, we propose a simple yet effective detector called LSDA (\underline{L}atent \underline{S}pace \underline{D}ata \underline{A}ugmentation), which is based on a heuristic idea: representations with a wider variety of forgeries should be able to learn a more generalizable decision boundary, thereby mitigating the overfitting of method-specific features (see Fig.~\ref{fig:toy}). Following this idea, we propose to enlarge the forgery space by constructing and simulating variations within and across forgery features in the latent space. This approach encompasses the acquisition of enriched, domain-specific features and the facilitation of smoother transitions between different forgery types, effectively bridging domain gaps. Our approach culminates in refining a binary classifier that leverages the distilled knowledge from the enhanced features, striving for a generalizable deepfake detector. Comprehensive experiments show that our proposed method is surprisingly effective and transcends state-of-the-art detectors across several widely used benchmarks.

Yexing Xu, Wei Feng, Shen Zhang et al.

Generating realistic and user-preferred advertisements is a key challenge in e-commerce. Existing approaches utilize multiple independent models driven by click-through-rate (CTR) to controllably create attractive image or text advertisements. However, their pipelines lack cross-modal perception and rely on CTR that only reflects average preferences. Therefore, we explore jointly generating personalized image-text advertisements from historical click behaviors. We first design a Unified Advertisement Generative model (Uni-AdGen) that employs a single autoregressive framework to produce both advertising images and texts. By incorporating a foreground perception module and instruction tuning, Uni-AdGen enhances the realism of the generated content. To further personalize advertisements, we equip Uni-AdGen with a coarse-to-fine preference understanding module that effectively captures user interests from noisy multimodal historical behaviors to drive personalized generation. Additionally, we construct the first large-scale Personalized Advertising image-text dataset (PAd1M) and introduce a Product Background Similarity (PBS) metric to facilitate training and evaluation. Extensive experiments show that our method outperforms baselines in general and personalized advertisement generation. Our project is available at https://github.com/JD-GenX/Uni-AdGen.

Zizhang Luo, Yuhao Luo, Youwei Xiao et al. · pku

Large language models are increasingly deployed as complex agentic systems that scale with task complexity. While prior work has extensively explored model- and system-level scaling, algorithm- and task-level scaling remain largely unaddressed, constraining the full potential of agentic systems. At the algorithm level, allocating additional inference-time computation can enhance workflow capacity but introduces cross-path redundancy: overlapping computations across multiple reasoning branches. At the task level, complex tasks can be decomposed into subproblems and delegated across multiple agents for improved scalability and parallelism. However, existing infrastructures' scheduling is unaware of the existence of multiple agents, missing opportunities to optimize resource allocation. We propose Hive, a multi-agent infrastructure that enables algorithm- and task-level scaling. Hive features a description frontend that captures per-agent behavior and supports test-time scaling algorithms. Leveraging this specification, our backend introduces two key mechanisms: Logits Cache that reuses intermediate logits across redundant sampling paths to mitigate cross-path redundancy at the algorithm level, and Agent-Aware Scheduling that efficiently allocates compute and KV-cache resources according to agent contributions at the task level. Experiments show that Logits Cache achieves an average speedup of $1.11\times$-$1.76\times$ for re-sampling, and Agent-Aware Scheduling reduces the hotspot miss rate by $33\%$-$51\%$.

Zizhang Luo, Yansong Xu, Runlin Guo et al. · pku

RTL program repair remains a critical bottleneck in hardware design and verification. Traditional automatic program repair (APR) methods rely on predefined templates and synthesis, limiting their bug coverage. Large language models (LLMs) and coding agents based on them offer flexibility but suffer from randomness and context corruption when handling long RTL code and waveforms. We present Clover, a neural-symbolic agentic harness that orchestrates RTL repair as a structured search over code manipulations to explore a validated solution for the bug. Recognizing that different repair operations favor distinct strategies, Clover dynamically dispatches tasks to specialized LLM agents or symbolic solvers. At its core, Clover introduces stochastic tree-of-thoughts, a test-time scaling mechanism that manages the main agent's context as a search tree, balancing exploration and exploitation for reliable outcomes. An RTL-specific toolbox further empowers agents to interact with the debugging environment. Evaluated on the RTL-repair benchmark, Clover fixes 96.8% of bugs within a fixed time limit, covering 94% and 63% more bugs than both pure traditional and LLM-based baselines, respectively, while achieving an average pass@1 rate of 87.5%, demonstrating high reliability and effectiveness.

Jikang Cheng, Zhiyuan Yan, Ying Zhang et al.

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 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 (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 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 Oriented Progressive Regularizor (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.

Yuhao Luo, Kehua Chen, Meixin Zhu

As a vital component in autonomous driving, accurate trajectory prediction effectively prevents traffic accidents and improves driving efficiency. To capture complex spatial-temporal dynamics and social interactions, recent studies developed models based on advanced deep-learning methods. On the other hand, recent studies have explored the use of deep generative models to further account for trajectory uncertainties. However, the current approaches demonstrating indeterminacy involve inefficient and time-consuming practices such as sampling from trained models. To fill this gap, we proposed a novel model named Graph Recurrent Attentive Neural Process (GRANP) for vehicle trajectory prediction while efficiently quantifying prediction uncertainty. In particular, GRANP contains an encoder with deterministic and latent paths, and a decoder for prediction. The encoder, including stacked Graph Attention Networks, LSTM and 1D convolutional layers, is employed to extract spatial-temporal relationships. The decoder is used to learn a latent distribution and thus quantify prediction uncertainty. To reveal the effectiveness of our model, we evaluate the performance of GRANP on the highD dataset. Extensive experiments show that GRANP achieves state-of-the-art results and can efficiently quantify uncertainties. Additionally, we undertake an intuitive case study that showcases the interpretability of the proposed approach. The code is available at https://github.com/joy-driven/GRANP.

Yuyang Zou, Youwei Xiao, Chenyun Yin et al.

Application-Specific Instruction-Set Processors (ASIPs) built on the RISC-V architecture offer specialization opportunities for various applications. Existing frameworks are largely designed around fixed instruction extension interfaces and rely on manual software adaptation. However, as emerging domains scale up in complexity, two major challenges arise. First, memory access remains a primary bottleneck as existing design flows lack architectural awareness of memory interfaces, leading to suboptimal interface selection and orchestration. Second, the semantic complexity of custom instruction extensions, characterized by non-trivial control logic and irregular memory behaviors, hinders the ability of conventional compilers to perform automated and comprehensive offloading. We present Aquas, a holistic hardware-software co-design framework built upon MLIR. Aquas proposes a memory interface model that jointly considers interface characteristics and cache effects, along with an interface-aware synthesis flow guided by this model that progressively optimizes the input specification and generates efficient hardware implementations. We also propose an e-graph-based retargetable compiler approach with a novel matching engine for efficient instruction mapping and offloading, enabling robust and effective utilization of custom instruction capabilities. Case studies across four diverse domains show that Aquas delivers substantial acceleration, achieving up to 15.61x speedup with 14.5% area overhead and zero frequency degradation, proving highly competitive in domain acceleration against more powerful general-purpose cores and vector extensions.

Yuhao Luo, Shiwei Ma, Mingjun Nie et al.

Cross-domain recommendation (CDR) is an important method to improve recommender system performance, especially when observations in target domains are sparse. However, most existing cross-domain recommendations fail to fully utilize the target domain's special features and are hard to be generalized to new domains. The designed network is complex and is not suitable for rapid industrial deployment. Our method introduces a two-step domain-aware cross-attention, extracting transferable features of the source domain from different granularity, which allows the efficient expression of both domain and user interests. In addition, we simplify the training process, and our model can be easily deployed on new domains. We conduct experiments on both public datasets and industrial datasets, and the experimental results demonstrate the effectiveness of our method. We have also deployed the model in an online advertising system and observed significant improvements in both Click-Through-Rate (CTR) and effective cost per mille (ECPM).

Zihao Sheng, Zilin Huang, Yansong Qu et al.

Multimodal large language models (MLLMs) have achieved remarkable progress across a range of vision-language tasks and demonstrate strong potential for traffic accident understanding. However, existing MLLMs in this domain primarily focus on coarse-grained image-level or video-level comprehension and often struggle to handle fine-grained visual details or localized scene components, limiting their applicability in complex accident scenarios. To address these limitations, we propose SafePLUG, a novel framework that empowers MLLMs with both Pixel-Level Understanding and temporal Grounding for comprehensive traffic accident analysis. SafePLUG supports both arbitrary-shaped visual prompts for region-aware question answering and pixel-level segmentation based on language instructions, while also enabling the recognition of temporally anchored events in traffic accident scenarios. To advance the development of MLLMs for traffic accident understanding, we curate a new dataset containing multimodal question-answer pairs centered on diverse accident scenarios, with detailed pixel-level annotations and temporal event boundaries. Experimental results show that SafePLUG achieves strong performance on multiple tasks, including region-based question answering, pixel-level segmentation, temporal event localization, and accident event understanding. These capabilities lay a foundation for fine-grained understanding of complex traffic scenes, with the potential to improve driving safety and enhance situational awareness in smart transportation systems. The code, dataset, and model checkpoints will be made publicly available at: https://zihaosheng.github.io/SafePLUG

Yuxuan Du, Zhendong Wang, Yuhao Luo et al.

The rapid emergence of multimodal deepfakes (visual and auditory content are manipulated in concert) undermines the reliability of existing detectors that rely solely on modality-specific artifacts or cross-modal inconsistencies. In this work, we first demonstrate that modality-specific forensic traces (e.g., face-swap artifacts or spectral distortions) and modality-shared semantic misalignments (e.g., lip-speech asynchrony) offer complementary evidence, and that neglecting either aspect limits detection performance. Existing approaches either naively fuse modality-specific features without reconciling their conflicting characteristics or focus predominantly on semantic misalignment at the expense of modality-specific fine-grained artifact cues. To address these shortcomings, we propose a general multimodal framework for video deepfake detection via Cross-Modal Alignment and Distillation (CAD). CAD comprises two core components: 1) Cross-modal alignment that identifies inconsistencies in high-level semantic synchronization (e.g., lip-speech mismatches); 2) Cross-modal distillation that mitigates feature conflicts during fusion while preserving modality-specific forensic traces (e.g., spectral distortions in synthetic audio). Extensive experiments on both multimodal and unimodal (e.g., image-only/video-only)deepfake benchmarks demonstrate that CAD significantly outperforms previous methods, validating the necessity of harmonious integration of multimodal complementary information.

Zilin Huang, Zihao Sheng, Zhengyang Wan et al.

Recent advances in autonomous system simulation platforms have significantly enhanced the safe and scalable testing of driving policies. However, existing simulators do not yet fully meet the needs of future transportation research-particularly in enabling effective human-AI collaboration and modeling socially-aware driving agents. This paper introduces Sky-Drive, a novel distributed multi-agent simulation platform that addresses these limitations through four key innovations: (a) a distributed architecture for synchronized simulation across multiple terminals; (b) a multi-modal human-in-the-loop framework integrating diverse sensors to collect rich behavioral data; (c) a human-AI collaboration mechanism supporting continuous and adaptive knowledge exchange; and (d) a digital twin framework for constructing high-fidelity virtual replicas of real-world transportation environments. Sky-Drive supports diverse applications such as autonomous vehicle-human road users interaction modeling, human-in-the-loop training, socially-aware reinforcement learning, personalized driving development, and customized scenario generation. Future extensions will incorporate foundation models for context-aware decision support and hardware-in-the-loop testing for real-world validation. By bridging scenario generation, data collection, algorithm training, and hardware integration, Sky-Drive has the potential to become a foundational platform for the next generation of human-centered and socially-aware autonomous transportation systems research. The demo video and code are available at:https://sky-lab-uw.github.io/Sky-Drive-website/

Yuhao Luo, Yuang Zhang, Kehua Chen et al.

Accurate pedestrian trajectory prediction is crucial for ensuring safety and efficiency in autonomous driving and human-robot interaction scenarios. Earlier studies primarily utilized sufficient observational data to predict future trajectories. However, in real-world scenarios, such as pedestrians suddenly emerging from blind spots, sufficient observational data is often unavailable (i.e. momentary trajectory), making accurate prediction challenging and increasing the risk of traffic accidents. Therefore, advancing research on pedestrian trajectory prediction under extreme scenarios is critical for enhancing traffic safety. In this work, we propose a novel framework termed Diffusion^2, tailored for momentary trajectory prediction. Diffusion^2 consists of two sequentially connected diffusion models: one for backward prediction, which generates unobserved historical trajectories, and the other for forward prediction, which forecasts future trajectories. Given that the generated unobserved historical trajectories may introduce additional noise, we propose a dual-head parameterization mechanism to estimate their aleatoric uncertainty and design a temporally adaptive noise module that dynamically modulates the noise scale in the forward diffusion process. Empirically, Diffusion^2 sets a new state-of-the-art in momentary trajectory prediction on ETH/UCY and Stanford Drone datasets.

Mingyue Yuan, Dong Yin, Jingwen Ding et al.

Person re-identification plays an important role in realistic video surveillance with increasing demand for public safety. In this paper, we propose a novel framework with rules of updating images for person re-identification in real-world surveillance system. First, Image Pool is generated by using mean-shift tracking method to automatically select video frame fragments of the target person. Second, features extracted from Image Pool by convolutional network work together to re-rank original ranking list of the main image and matching results will be generated. In addition, updating rules are designed for replacing images in Image Pool when a new image satiating with our updating critical formula in video system. These rules fall into two categories: if the new image is from the same camera as the previous updated image, it will replace one of assist images; otherwise, it will replace the main image directly. Experiments are conduced on Market-1501, iLIDS-VID and PRID-2011 and our ITSD datasets to validate that our framework outperforms on rank-1 accuracy and mAP for person re-identification. Furthermore, the update ability of our framework provides consistently remarkable accuracy rate in real-world surveillance system.