arXiv subjects

Zheming Yang

Publications and source records attributed to Zheming Yang.

At least 19 recordsLinked to original sources

arXiv2026

Pro-Router: Token-Aware Progressive Model Routing with Adaptive Edge-Cloud Collaboration for Efficient Multimodal LLM Inference

The remarkable performance of multimodal large language models (MLLMs) comes at the cost of substantial computational overhead, posing significant challenges to real-time deployment and cost effectiveness. Existing model routing approaches either decide from coarse request-level features alone or spend one or several extra language model passes to inspect the generated response, leaving the token-level uncertainty signals that emerge during generation unused. To address these limitations, we propose Pro-Router, a token-aware progressive model routing method with adaptive edge-cloud collaboration for efficient multimodal LLM inference. Pro-Router employs a two-stage progressive decision mechanism. First, a lightweight prompt pre-scorer module performs rapid pre-screening before token generation begins, guiding apparently simple requests to small models. Second, a token-aware verifier reads the sampling probability distribution of each token the small model generates, estimating the model's confidence in its own output to determine, per request, whether the answer ships or escalates to the cloud-based high-precision model. Furthermore, we design an adaptive edge-cloud serving pipeline that sizes every dispatch to each device's measured service rate, so both the edge and the cloud tiers stay fully utilized without manual parameter tuning and are not impacted by the network latency. Extensive experiments on multiple multimodal benchmark datasets and models demonstrate the effectiveness of Pro-Router. Compared to other methods, it achieves the highest routing accuracy and improves routing speed by more than 10x. Its serving pipeline also reaches more than 75% higher end-to-end throughput than the existing model routing pipeline. Our code is available at https://github.com/xinyuangui2/pro-router.

cs.AI↗

arXiv2026

When Is Thinking Enough? Early Exit via Sufficiency Assessment for Efficient Reasoning

Large reasoning models (LRMs) have achieved remarkable performance in complex reasoning tasks, driven by their powerful inference-time scaling capability. However, LRMs often suffer from overthinking, which results in substantial computational redundancy and significantly reduces efficiency. Early-exit methods aim to mitigate this issue by terminating reasoning once sufficient evidence has been generated, yet existing approaches mostly rely on handcrafted or empirical indicators that are unreliable and impractical. In this work, we introduce Dynamic Thought Sufficiency in Reasoning (DTSR), a novel framework for efficient reasoning that enables the model to dynamically assess the sufficiency of its chain-of-thought (CoT) and determine the optimal point for early exit. Inspired by human metacognition, DTSR operates in two stages: (1) Reflection Signal Monitoring, which identifies reflection signals as potential cues for early exit, and (2) Thought Sufficiency Check, which evaluates whether the current CoT is sufficient to derive the final answer. Experimental results on the Qwen3 models show that DTSR reduces reasoning length by 28.9%-34.9% with minimal performance loss, effectively mitigating overthinking. We further discuss overconfidence in LRMs and self-evaluation paradigms, providing valuable insights for early-exit reasoning.

cs.CL↗

arXiv2026

DAT: Dual-Aware Adaptive Transmission for Efficient Multimodal LLM Inference in Edge-Cloud Systems

Multimodal large language models (MLLMs) have shown strong capability in semantic understanding and visual reasoning, yet their use on continuous video streams in bandwidth-constrained edge-cloud systems incurs prohibitive computation and communication overhead and hinders low-latency alerting and effective visual evidence delivery. To address this challenge, we propose DAT to achieve high-quality semantic generation, low-latency event alerting, and effective visual evidence supplementation. To reduce unnecessary deep reasoning costs, we propose a collaborative small-large model cascade. A lightweight edge-side small model acts as a gating module to filter non-target-event frames and perform object detection, triggering MLLM inference only for suspicious frames. Building on this, we introduce an efficient fine-tuning strategy with visual guidance and semantic prompting, which improves structured event understanding, object detection, and output consistency. To ensure low-latency semantic alerting and effective visual evidence supplementation under bandwidth constraints, we further devise a semantics and bandwidth-aware multi-stream adaptive transmission optimization method. Experimental results show that DAT achieves 98.83% recognition accuracy and 100% output consistency. Under severe congestion, it reduces weighted semantic alert delay by up to 77.5% and delivers 98.33% of visual evidence within 0.5 s, demonstrating the effectiveness of jointly optimizing cascade inference and elastic transmission.

cs.MM↗

arXiv2026

MSAO: Adaptive Modality Sparsity-Aware Offloading with Edge-Cloud Collaboration for Efficient Multimodal LLM Inference

Multimodal large language models (MLLMs) enable powerful cross-modal reasoning capabilities but impose substantial computational and latency burdens, posing critical challenges for deployment on resource-constrained edge devices. In this paper, we propose MSAO, an adaptive modality sparsity-aware offloading framework with edge-cloud collaboration for efficient MLLM Inference. First, a lightweight heterogeneous modality-aware via fine-grained sparsity module performs spatial-temporal-modal joint analysis to compute the Modality Activation Sparsity (MAS) metric, which quantifies the necessity of each modality with minimal computational overhead. Second, an adaptive speculative edge-cloud collaborative offloading mechanism dynamically schedules workloads between edge and cloud based on the derived MAS scores and real-time system states, leveraging confidence-guided speculative execution to hide communication latency. Extensive experiments on VQAv2 and MMBench benchmarks demonstrate that MSAO achieves a 30% reduction in end-to-end latency and 30%-65% decrease in resource overhead, while delivering a throughput improvement of 1.5x to 2.3x compared to traditional approaches, all without compromising competitive accuracy.

cs.DC↗

arXiv2026

R2E-VID: Two-Stage Robust Routing via Temporal Gating for Elastic Edge-Cloud Video Inference

With the rapid growth of large-scale video analytics applications, edge-cloud collaborative systems have become the dominant paradigm for real-time inference. However, existing approaches often fail to dynamically adapt to heterogeneous video content and fluctuating resource conditions, resulting in suboptimal routing efficiency and high computational costs. In this paper, we propose R2E-VID, a two-stage robust routing framework via temporal gating for elastic edge-cloud video inference. In the first stage, R2E-VID introduces a temporal gating mechanism that models the temporal consistency and motion dynamics of incoming video streams to predict the optimal routing pattern for each segment. This enables adaptive partitioning of inference workloads between edge and cloud nodes, achieving fine-grained spatiotemporal elasticity. In the second stage, a robust routing optimization module refines the allocation through multi-model adaptation, jointly minimizing inference delay and resource consumption under dynamic network and workload variations. Extensive experiments on public datasets demonstrate that R2E-VID achieves up to 60% reduction in overall cost compared to cloud-centric baselines, and delivers 35-45% lower delay while improving inference accuracy by 2-7% over state-of-the-art edge-cloud solutions.

cs.NI↗

arXiv2026

Not All Negative Samples Are Equal: LLMs Learn Better from Plausible Reasoning

Learning from negative samples holds great promise for improving Large Language Model (LLM) reasoning capability, yet existing methods treat all incorrect responses as equally informative, overlooking the crucial role of sample quality. To address this, we propose Plausible Negative Samples (PNS), a method that synthesizes high-quality negative samples exhibiting expected format and structural coherence while ultimately yielding incorrect answers. PNS trains a dedicated model via reverse reinforcement learning (RL) guided by a composite reward combining format compliance, accuracy inversion, reward model assessment, and chain-of-thought evaluation, generating responses nearly indistinguishable from correct solutions. We further validate PNS as a plug-and-play data source for preference optimization across three backbone models on seven mathematical reasoning benchmarks. Results demonstrate that PNS consistently outperforms other negative sample synthesis methods, achieving an average improvement of 2.03% over RL-trained models.

cs.LG↗

arXiv2026

From Atoms to Chains: Divergence-Guided Reasoning Curriculum for Unlabeled LLM Domain Adaptation

Adapting Large Language Models (LLMs) to specialized domains without human-annotated data is a crucial yet formidable challenge. Widely adopted knowledge distillation methods often devolve into coarse-grained mimicry, where the student model inefficiently targets its own weaknesses and risks inheriting the teacher's reasoning flaws. This exposes a critical pedagogical dilemma: how to devise a reliable curriculum when the teacher itself is not an infallible expert. Our work resolves this by capitalizing on a key insight: while LLMs may exhibit fallibility in complex, holistic reasoning, they often exhibit high fidelity on focused, atomic sub-problems. Based on this, we propose Divergence-Guided Reasoning Curriculum (DGRC), which constructs a learning path from atomic knowledge to reasoning chains by dynamically deriving two complementary curricula from disagreements in reasoning pathways. When a student and teacher produce conflicting results, DGRC directs the teacher to perform a diagnostic analysis: it analyzes both reasoning paths to formulate atomic queries that target the specific points of divergence, and then self-answers these queries to create high-confidence atomic question-answer pairs. These pairs then serve a dual purpose: (1) providing an atomic curriculum to rectify the student's knowledge gaps, and (2) serving as factual criteria to filter the teacher's original reasoning chains, yielding a verified CoT curriculum that teaches the student how to integrate atomic knowledge into complete reasoning paths. Experiments across the medical and legal domains on student models of various sizes demonstrate the effectiveness of our DGRC framework. Notably, our method achieves a 7.76% relative improvement for the 1.5B student model in the medical domain over strong unlabeled baseline.

cs.LG↗

arXiv2026

Mimic Human Cognition, Master Multi-Image Reasoning: A Meta-Action Framework for Enhanced Visual Understanding

While Multimodal Large Language Models (MLLMs) excel at single-image understanding, they exhibit significantly degraded performance in multi-image reasoning scenarios. Multi-image reasoning presents fundamental challenges including complex inter-relationships between images and scattered critical information across image sets. Inspired by human cognitive processes, we propose a Cognition-Inspired Meta-Action Framework (CINEMA), which decomposes multi-image reasoning into five structured meta-actions: Global, Focus, Hint, Think, and Answer, explicitly modeling the sequential cognitive steps humans naturally employ. For cold-start training, we introduce a Retrieval-Based Tree Sampling strategy that generates high-quality meta-action trajectories to bootstrap the model with reasoning patterns. During reinforcement learning, we adopt a two-stage paradigm: an exploration phase with Diversity-Preserving Strategy to avoid entropy collapse, followed by an annealed exploitation phase with DAPO to gradually strengthen exploitation. To train our model, we construct a dataset of 56k cold-start and 58k reinforcement learning instances spanning multi-image, multi-frame, and single-image tasks. We conduct extensive evaluations on multi-image reasoning benchmarks, video understanding benchmarks, and single-image benchmarks, achieving competitive state-of-the-art performance on several key benchmarks. Our model surpasses GPT-4o on the MUIR and MVMath benchmarks and notably outperforms specialized video reasoning models on video understanding benchmarks, demonstrating the effectiveness and generalizability of our human cognition-inspired reasoning framework.

cs.CV↗

arXiv2026

ThinkDrive: Chain-of-Thought Guided Progressive Reinforcement Learning Fine-Tuning for Autonomous Driving

With the rapid advancement of large language models (LLMs) technologies, their application in the domain of autonomous driving has become increasingly widespread. However, existing methods suffer from unstructured reasoning, poor generalization, and misalignment with human driving intent. While Chain-of-Thought (CoT) reasoning enhances decision transparency, conventional supervised fine-tuning (SFT) fails to fully exploit its potential, and reinforcement learning (RL) approaches face instability and suboptimal reasoning depth. We propose ThinkDrive, a CoT guided progressive RL fine-tuning framework for autonomous driving that synergizes explicit reasoning with difficulty-aware adaptive policy optimization. Our method employs a two-stage training strategy. First, we perform SFT using CoT explanations. Then, we apply progressive RL with a difficulty-aware adaptive policy optimizer that dynamically adjusts learning intensity based on sample complexity. We evaluate our approach on a public dataset. The results show that ThinkDrive outperforms strong RL baselines by 1.45%, 1.95%, and 1.01% on exam, easy-exam, and accuracy, respectively. Moreover, a 2B-parameter model trained with our method surpasses the much larger GPT-4o by 3.28% on the exam metric.

cs.AI↗

arXiv2026

AIVD: Adaptive Edge-Cloud Collaboration for Accurate and Efficient Industrial Visual Detection

Multimodal large language models (MLLMs) demonstrate exceptional capabilities in semantic understanding and visual reasoning, yet they still face challenges in precise object localization and resource-constrained edge-cloud deployment. To address this, this paper proposes the AIVD framework, which achieves unified precise localization and high-quality semantic generation through the collaboration between lightweight edge detectors and cloud-based MLLMs. To enhance the cloud MLLM's robustness against edge cropped-box noise and scenario variations, we design an efficient fine-tuning strategy with visual-semantic collaborative augmentation, significantly improving classification accuracy and semantic consistency. Furthermore, to maintain high throughput and low latency across heterogeneous edge devices and dynamic network conditions, we propose a heterogeneous resource-aware dynamic scheduling algorithm. Experimental results demonstrate that AIVD substantially reduces resource consumption while improving MLLM classification performance and semantic generation quality. The proposed scheduling strategy also achieves higher throughput and lower latency across diverse scenarios.

cs.CV↗

arXiv2026

SearchAttack: Red-Teaming LLMs against Knowledge-to-Action Threats under Online Web Search

Recently, people have suffered from LLM hallucination and have become increasingly aware of the reliability gap of LLMs in open and knowledge-intensive tasks. As a result, they have increasingly turned to search-augmented LLMs to mitigate this issue. However, LLM-driven search also becomes an attractive target for misuse. Once the returned content directly contains targeted, ready-to-use harmful instructions or takeaways for users, it becomes difficult to withdraw or undo such exposure. To investigate LLMs' unsafe search behavior issues, we first propose \textbf{\textit{SearchAttack}} for red-teaming, which (1) rephrases harmful semantics via dense and benign knowledge to evade direct in-context decoding, thus eliciting unsafe information retrieval, (2) stress-tests LLMs' reward-chasing bias by steering them to synthesize unsafe retrieved content. We also curate an emergent, domain-specific illicit activity benchmark for search-based threat assessment, and introduce a fact-checking framework to ground and quantify harm in both offline and online attack settings. Extensive experiments are conducted to red-team the search-augmented LLMs for responsible vulnerability assessment. Empirically, SearchAttack demonstrates strong effectiveness in attacking these systems. We also find that LLMs without web search can still be steered into harmful content output due to their information-seeking stereotypical behaviors.

cs.CL↗

arXiv2025

MMLongCite: A Benchmark for Evaluating Fidelity of Long-Context Vision-Language Models

The rapid advancement of large vision language models (LVLMs) has led to a significant expansion of their context windows. However, an extended context window does not guarantee the effective utilization of the context, posing a critical challenge for real-world applications. Current evaluations of such long-context faithfulness are predominantly focused on the text-only domain, while multimodal assessments remain limited to short contexts. To bridge this gap, we introduce MMLongCite, a comprehensive benchmark designed to evaluate the fidelity of LVLMs in long-context scenarios. MMLongCite comprises 8 distinct tasks spanning 6 context length intervals and incorporates diverse modalities, including text, images, and videos. Our evaluation of state-of-the-art LVLMs reveals their limited faithfulness in handling long multimodal contexts. Furthermore, we provide an in-depth analysis of how context length and the position of crucial content affect the faithfulness of these models.

cs.CV↗

arXiv2025

Logo-VGR: Visual Grounded Reasoning for Open-world Logo Recognition

Recent advances in multimodal large language models (MLLMs) have been primarily evaluated on general-purpose benchmarks, while their applications in domain-specific scenarios, such as intelligent product moderation, remain underexplored. To address this gap, we introduce an open-world logo recognition benchmark, a core challenge in product moderation. Unlike traditional logo recognition methods that rely on memorizing representations of tens of thousands of brands-an impractical approach in real-world settings-our proposed method, Logo-VGR, enables generalization to large-scale brand recognition with supervision from only a small subset of brands. Specifically, we reformulate logo recognition as a comparison-based task, requiring the model to match product images with candidate logos rather than directly generating brand labels. We further observe that existing models tend to overfit by memorizing brand distributions instead of learning robust multimodal reasoning, which results in poor performance on unseen brands. To overcome this limitation, Logo-VGR introduces a new paradigm of domain-specific multimodal reasoning: Logo Perception Grounding injects domain knowledge, and Logo-Guided Visual Grounded Reasoning enhances the model's reasoning capability. Experimental results show that Logo-VGR outperforms strong baselines by nearly 10 points in OOD settings, demonstrating superior generalization.

cs.CV↗

arXiv2025

Adaptive Guidance Semantically Enhanced via Multimodal LLM for Edge-Cloud Object Detection

Traditional object detection methods face performance degradation challenges in complex scenarios such as low-light conditions and heavy occlusions due to a lack of high-level semantic understanding. To address this, this paper proposes an adaptive guidance-based semantic enhancement edge-cloud collaborative object detection method leveraging Multimodal Large Language Models (MLLM), achieving an effective balance between accuracy and efficiency. Specifically, the method first employs instruction fine-tuning to enable the MLLM to generate structured scene descriptions. It then designs an adaptive mapping mechanism that dynamically converts semantic information into parameter adjustment signals for edge detectors, achieving real-time semantic enhancement. Within an edge-cloud collaborative inference framework, the system automatically selects between invoking cloud-based semantic guidance or directly outputting edge detection results based on confidence scores. Experiments demonstrate that the proposed method effectively enhances detection accuracy and efficiency in complex scenes. Specifically, it can reduce latency by over 79% and computational cost by 70% in low-light and highly occluded scenes while maintaining accuracy.

cs.CV↗

arXiv2025

MoA-Off: Adaptive Heterogeneous Modality-Aware Offloading with Edge-Cloud Collaboration for Efficient Multimodal LLM Inference

Multimodal large language models (MLLMs) enable powerful cross-modal inference but impose significant computational and latency burdens, posing severe challenges for deployment in resource-constrained environments. In this paper, we propose MoA-Off, an adaptive heterogeneous modality-aware offloading framework with edge-cloud collaboration for efficient MLLM inference. MoA-Off introduces a lightweight heterogeneous modality-aware module that estimates the complexity of heterogeneous inputs through multi-dimensional feature analysis. Then, an adaptive edge-cloud collaborative offloading strategy is proposed that dynamically schedules workloads between edge and cloud based on modality-aware complexity scores and real-time system states. The experimental results demonstrate that MoA-Off can achieve over 30% reduction in latency and 30%-65% decrease in resource overhead while maintaining competitive accuracy compared to traditional approaches.

cs.DC↗

arXiv2025

SAEC: Scene-Aware Enhanced Edge-Cloud Collaborative Industrial Vision Inspection with Multimodal LLM

Industrial vision inspection requires high accuracy under stringent resource constraints, yet existing approaches face a fundamental trade-off. Multimodal LLMs (MLLMs) deliver strong reasoning capabilities but incur prohibitive computational costs, while lightweight edge models often fail on complex cases. In this paper, we present SAEC, a scene-aware enhanced edge-cloud collaborative industrial vision inspection framework with MLLM. The framework is composed of three synergistic components: (1) Efficient MLLM Fine-Tuning for Complex Defect Inspection, (2) Lightweight Multiscale Scene-Complexity Estimation, and (3) Adaptive Edge-Cloud Scheduler. Together, these modules enable robust defect detection by tailoring multimodal reasoning to scene complexity and dynamically balancing computation between edge and cloud resources. Experimental results on MVTec AD and KSDD2 datasets demonstrate that SAEC attains 85.11% and 82.72% accuracy, surpassing Qwen by 22.1% and 20.8%, and LLaVA by 33.3% and 31.6%. It also reduces runtime by up to 22.4% and cuts energy per correct decision by 40%-74%. The code is available at https://github.com/YuHao-Tian/SAEC.

cs.CV↗

arXiv2025

CURE: Critical-Token-Guided Re-Concatenation for Entropy-Collapse Prevention

Recent advances in Reinforcement Learning with Verified Reward (RLVR) have driven the emergence of more sophisticated cognitive behaviors in large language models (LLMs), thereby enhancing their reasoning capabilities. However, in prior RLVR pipelines, the repeated use of static initial-state sampling drawn exactly from the dataset distribution during each sampling phase produced overly deterministic, low diversity model behavior, which manifested as rapid entropy collapse and hindered sustained performance gains during prolonged training. To address this issue, we introduce CURE (Critical-token-gUided Re concatenation for Entropy-collapse prevention), a two-stage framework that balances exploration and exploitation. Specifically, in the first stage, to deliberately steer the model toward novel yet coherent contexts, we re-generate at high-entropy critical tokens and jointly optimize the original and the branched trajectories. The further comparison with vanilla DAPO shows that the regeneration process achieves a better performance on math reasoning tasks while sustaining a high-level entropy degree for exploration. In the second stage, we continue training with static initial-state sampling by DAPO, intentionally placing the model in a familiar state to gradually strengthen exploitation. Extensive experiments on Qwen-2.5-Math-7B show that, compared to other RLVR methods, CURE achieves a 5% performance gain across six math benchmarks, establishing state-of-the-art performance in both entropy and accuracy. A series of experiments further validate the effectiveness of our approach. Code is available at https://github.com/bytedance/CURE.

cs.LG↗

arXiv2025

EC2MoE: Adaptive End-Cloud Pipeline Collaboration Enabling Scalable Mixture-of-Experts Inference

The Mixture-of-Experts (MoE) paradigm has emerged as a promising solution to scale up model capacity while maintaining inference efficiency. However, deploying MoE models across heterogeneous end-cloud environments poses new challenges in expert scheduling, communication overhead, and resource heterogeneity. In this paper, we propose EC2MoE, an adaptive framework for scalable MoE inference via end-cloud pipeline collaboration. First, we design a hardware-aware lightweight group gate network that enhances expert selection and computational efficiency. By incorporating a hardware-aware local expert selection mechanism, the system adaptively filters candidate experts based on real-time device profiles. A lightweight group gate module then integrates local and global gating outputs to achieve high-quality expert routing with minimal overhead. Second, we develop a pipeline optimization mechanism based on endcloud collaboration to accelerate MoE inference. This includes an encoder-decoder structure based on low-rank compression, which reduces transmission and computation costs. And a route-aware heuristic pipeline scheduling algorithm that dynamically allocates inference stages across devices according to workload and network topology. Extensive experiments show that EC2MoE can increase throughput by 2.2x to 5.1x and reduce end-to-end latency by 53% to 67% while maintaining high accuracy compared to state-of-the-art methods. It also maintains good scalability under dynamic load and network environments.

cs.DC↗