arXiv subjects

Jiajie Huang

Publications and source records attributed to Jiajie Huang.

10 recordsLinked to original sources

arXiv2026

WA-JEPA: Rethinking the Video JEPA Paradigm for World-Action Modeling in Autonomous Driving

Video Joint Embedding Predictive Architecture (V-JEPA) learns powerful spatiotemporal representations from video through self-supervised latent feature prediction. However, V-JEPA is built around random-mask completion and deterministic regression, making it fundamentally ill-suited for autonomous driving planning that demands future-directed prediction tightly coupled with action. To address this, we rethink the V-JEPA paradigm and present WA-JEPA, a V-JEPA-native world-action model designed for autonomous driving planning. Instead of random spatiotemporal masking, WA-JEPA employs hybrid future-masked pre-training, where the model infers future latents from observed context. Departing from deterministic regression, we recast future prediction as conditional flow matching over latent futures, which substantially improves the model's ability to generate plausible future latents for downstream planning. Finally, a joint future-action predictor is proposed to denoise future scene tokens and ego trajectories together in a unified spatiotemporal latent space, allowing action supervision to directly shape planning-relevant world representations. Pre-trained on nuPlan videos and fine-tuned on NAVSIM, WA-JEPA reaches 91.7 EPDMS on NAVSIM-v2, surpassing the strongest end-to-end and world-action baselines by 1.6 and 1.3 EPDMS, and, without HUGSIM-specific fine-tuning, attains the best HD-Score of 0.4462 on the closed-loop HUGSIM benchmark under the same evaluation protocol. These results validate V-JEPA-native world-action modeling as a powerful and scalable paradigm for autonomous driving planning. Code is available at https://github.com/AFARI-Research/WA-JEPA.

cs.CV↗

arXiv2026

CoWorld-VLA: Thinking in a Multi-Expert World Model for Autonomous Driving

Vision-Language-Action (VLA) models have emerged as a promising paradigm for end-to-end autonomous driving. However, existing reasoning mechanisms still struggle to provide planning-oriented intermediate representations: textual Chain-of-Thought (CoT) fails to preserve continuous spatiotemporal structure, while latent world reasoning remains difficult to use as a direct condition for action generation. In this paper, we propose CoWorld-VLA, a multi-expert world reasoning framework for autonomous driving, where world representations serve as explicit conditions to guide action planning. CoWorld-VLA extracts complementary world information through multi-source supervision and encodes it into expert tokens within the VLA, thereby providing planner-accessible conditioning signals. Specifically, we construct four types of tokens: semantic interaction, geometric structure, dynamic evolution, and ego trajectory tokens, which respectively model interaction intent, spatial structure, future temporal dynamics, and behavioral goals. During action generation, CoWorld-VLA employs a diffusion-based hierarchical multi-expert fusion planner, which is coupled with scene context throughout the joint denoising process to generate continuous ego trajectories. Experiments on NAVSIM v1 & v2 demonstrate future-scene modeling capability and strong planning performance, including collision avoidance and trajectory accuracy. Ablation studies further validate the complementarity of expert tokens and their effectiveness as planning conditions for action generation. Code will be available at https://github.com/AFARI-Research/CoWorld-VLA.

cs.CV↗

arXiv2026

EvoStreaming: Your Offline Video Model Is a Natively Streaming Assistant

Streaming video understanding demands more than watching longer videos: assistants must decide when to speak in real time, balancing responsiveness against verbosity. Yet most video-language models (VideoLLMs) are trained for offline inference, and existing streaming benchmarks externalize this timing decision to the evaluator. We address this gap with RealStreamEval, a frame-level multi-turn evaluation protocol that exposes models to sequential observations and penalizes unnecessary responses. Under this protocol, we observed that strong offline VideoLLMs retain useful visual understanding but lack an interaction policy for deciding when to respond. Motivated by this observation, we propose EvoStreaming, a self-evolved streaming adaptation framework in which the base model itself acts as data generator, relevance annotator, and roll-out policy to synthesize streaming trajectories without external supervision. With only $1{,}000$ self-generated samples ($139\times$ less than the leading streaming instruction-tuning approach) and no architectural changes, EvoStreaming consistently improves the overall RealStreamEval score by up to $10.8$ points across five open VideoLLM backbones (Qwen2/2.5/3-VL, InternVL-3.5, MiniCPM-V4.5) while largely preserving offline video performance. These results suggest that data-efficient interaction tuning is a practical path for adapting existing VideoLLMs to streaming assistants.

cs.CV↗

arXiv2026

MedLA: A Logic-Driven Multi-Agent Framework for Complex Medical Reasoning with Large Language Models

Answering complex medical questions requires not only domain expertise and patient-specific information, but also structured and multi-perspective reasoning. Existing multi-agent approaches often rely on fixed roles or shallow interaction prompts, limiting their ability to detect and resolve fine-grained logical inconsistencies. To address this, we propose \textsc{MedLA}, a logic-driven multi-agent framework built on large language models. Each agent organizes its reasoning process into an explicit logical tree based on syllogistic triads (major premise, minor premise, and conclusion), enabling transparent inference and premise-level alignment. Agents engage in a multi-round, graph-guided discussion to compare and iteratively refine their logic trees, achieving consensus through error correction and contradiction resolution. We demonstrate that \textsc{MedLA} consistently outperforms both static role-based systems and single-agent baselines on challenging benchmarks such as MedDDx and standard medical QA tasks. Furthermore, \textsc{MedLA} scales effectively across both open-source and commercial LLM backbones, achieving state-of-the-art performance and offering a generalizable paradigm for trustworthy medical reasoning.

cs.AI↗

arXiv2026

From Simulation to Reality: Practical Deep Reinforcement Learning-based Link Adaptation for Cellular Networks

Link Adaptation (LA) that dynamically adjusts the Modulation and Coding Schemes (MCS) to accommodate time-varying channels is crucial and challenging in cellular networks. Deep reinforcement learning (DRL)-based LA that learns to make decision through the interaction with the environment is a promising approach to improve throughput. However, existing DRL-based LA algorithms are typically evaluated in simplified simulation environments, neglecting practical issues such as ACK/NACK feedback delay, retransmission and parallel hybrid automatic repeat request (HARQ). Moreover, these algorithms overlook the impact of DRL execution latency, which can significantly degrade system performance. To address these challenges, we propose Decoupling-DQN (DC-DQN), a new DRL framework that separates traditional DRL's coupled training and inference processes into two modules based on Deep Q Networks (DQN): a real-time inference module and an out-of-decision-loop training module. Based on this framework, we introduce a novel DRL-based LA algorithm, DC-DQN-LA. The algorithm incorporates practical considerations by designing state, action, and reward functions that account for feedback delays, parallel HARQ, and retransmissions. We implemented a prototype using USRP software-defined radios and srsRAN software. Experimental results demonstrate that DC-DQN-LA improves throughput by 40\% to 70\% in mobile scenario compared with baseline LA algorithms, while maintaining comparable block error rates, and can quickly adapt to environment changes in mobile-to-static scenario. These results highlight the efficiency and practicality of the proposed DRL-based LA algorithm.

cs.NI↗

arXiv2025

Adaptive Query Prompting for Multi-Domain Landmark Detection

Medical landmark detection is crucial in various medical imaging modalities and procedures. Although deep learning-based methods have achieve promising performance, they are mostly designed for specific anatomical regions or tasks. In this work, we propose a universal model for multi-domain landmark detection by leveraging transformer architecture and developing a prompting component, named as Adaptive Query Prompting (AQP). Instead of embedding additional modules in the backbone network, we design a separate module to generate prompts that can be effectively extended to any other transformer network. In our proposed AQP, prompts are learnable parameters maintained in a memory space called prompt pool. The central idea is to keep the backbone frozen and then optimize prompts to instruct the model inference process. Furthermore, we employ a lightweight decoder to decode landmarks from the extracted features, namely Light-MLD. Thanks to the lightweight nature of the decoder and AQP, we can handle multiple datasets by sharing the backbone encoder and then only perform partial parameter tuning without incurring much additional cost. It has the potential to be extended to more landmark detection tasks. We conduct experiments on three widely used X-ray datasets for different medical landmark detection tasks. Our proposed Light-MLD coupled with AQP achieves SOTA performance on many metrics even without the use of elaborate structural designs or complex frameworks.

cs.CV↗

arXiv2025

Credal Transformer: A Principled Approach for Quantifying and Mitigating Hallucinations in Large Language Models

Large Language Models (LLMs) hallucinate, generating factually incorrect yet confident assertions. We argue this stems from the Transformer's Softmax function, which creates "Artificial Certainty" by collapsing ambiguous attention scores into a single probability distribution, discarding uncertainty information at each layer. To fix this, we introduce the Credal Transformer, which replaces standard attention with a Credal Attention Mechanism (CAM) based on evidential theory. CAM produces a "credal set" (a set of distributions) instead of a single attention vector, with the set's size directly measuring model uncertainty. We implement this by re-conceptualizing attention scores as evidence masses for a Dirichlet distribution: sufficient evidence recovers standard attention, while insufficient evidence yields a diffuse distribution, representing ambiguity. Empirically, the Credal Transformer identifies out-of-distribution inputs, quantifies ambiguity, and significantly reduces confident errors on unanswerable questions by abstaining. Our contribution is a new architecture to mitigate hallucinations and a design paradigm that integrates uncertainty quantification directly into the model, providing a foundation for more reliable AI.

cs.CL↗

arXiv2025

AudioMarathon: A Comprehensive Benchmark for Long-Context Audio Understanding and Efficiency in Audio LLMs

Processing long-form audio is a major challenge for Large Audio Language models (LALMs). These models struggle with the quadratic cost of attention ($O(N^2)$) and with modeling long-range temporal dependencies. Existing audio benchmarks are built mostly from short clips and do not evaluate models in realistic long context settings. To address this gap, we introduce AudioMarathon, a benchmark designed to evaluate both understanding and inference efficiency on long-form audio. AudioMarathon provides a diverse set of tasks built upon three pillars: long-context audio inputs with durations ranging from 90.0 to 300.0 seconds, which correspond to encoded sequences of 2,250 to 7,500 audio tokens, respectively, full domain coverage across speech, sound, and music, and complex reasoning that requires multi-hop inference. We evaluate state-of-the-art LALMs and observe clear performance drops as audio length grows. We also study acceleration techniques and analyze the trade-offs of token pruning and KV cache eviction. The results show large gaps across current LALMs and highlight the need for better temporal reasoning and memory-efficient architectures. We believe AudioMarathon will drive the audio and multimodal research community to develop more advanced audio understanding models capable of solving complex audio tasks.

cs.SD↗

arXiv2025

Age-Energy Analysis in Multi-Source Systems with Wake-up Control and Packet Management

In recent years, there has been an increasing focus on real-time mobile applications, such as news updates and weather forecast. In these applications, data freshness is of significant importance, which can be measured by age-of-synchronization (AoS). At the same time, the reduction of carbon emission is increasingly required by the communication operators. Thus, how to reduce energy consumption while keeping the data fresh becomes a matter of concern. In this paper, we study the age-energy trade-off in a multi-source single-server system, where the server can turn to sleep mode to save energy. We adopt the stochastic hybrid system (SHS) method to analyze the average AoS and power consumption with three wake-up policies including N-policy, single-sleep policy and multi-sleep policy, and three packet preemption strategies, including Last-Come-First-Serve with preemption-in-Service (LCFS-S), LCFS with preemption-only-in-Waiting (LCFS-W), and LCFS with preemption-and-Queueing (LCFS-Q). The trade-off performance is analyzed via both closed-form expressions and numerical simulations. It is found that N-policy attains the best trade-off performance among all three sleep policies. Among packet management strategies, LCFS-S is suitable for scenarios with high requirements on energy saving and small arrival rate difference between sources. LCFS-Q is suitable for scenarios with high requirements on information freshness and large arrival rate difference between sources.

cs.IT↗

arXiv2023

Age-Energy Trade-off in Status Update System with Wake-up Control

In the status update system, the freshness of information is drawing more and more attention. To measure the freshness of the data, age-of-synchronization (AoS) is introduced. Since many communication devices are energy-constrained, how to reduce energy consumption while keeping the timely response of data needs to be carefully addressed. In this paper, we study the age-energy trade-off problem in a single-source single-server scenario. We assume the server enters a low-power sleep state when idle to save energy and consider three wake-up policies. We adopt the stochastic hybrid system (SHS) method to analyze the average AoS and average energy consumption under different policies. The age-energy trade-off relationship under different parameters is illustrated by numerical results.

cs.IT↗