Recent advances in large language models (LLMs) have accelerated progress toward artificial general intelligence, with inference-time scaling emerging as a key technique. Contemporary approaches leverage either sequential reasoning (iteratively extending chains of thought) or parallel reasoning (generating multiple solutions simultaneously) to scale inference. However, both paradigms face fundamental limitations: sequential scaling typically relies on arbitrary token budgets for termination, leading to inefficiency or premature cutoff; while parallel scaling often lacks coordination among parallel branches and requires intrusive fine-tuning to perform effectively. In light of these challenges, we aim to design a flexible test-time collaborative inference framework that exploits the complementary strengths of both sequential and parallel reasoning paradigms. Towards this goal, the core challenge lies in developing an efficient and accurate intrinsic quality metric to assess model responses during collaborative inference, enabling dynamic control and early termination of the reasoning trace. To address this challenge, we introduce semantic entropy (SE), which quantifies the semantic diversity of parallel model responses and serves as a robust indicator of reasoning quality due to its strong negative correlation with accuracy...
The rapid adoption of generative AI (GenAI), particularly Large Language Models (LLMs), has exposed critical limitations of cloud-centric deployments, including latency, cost, and privacy concerns. Meanwhile, Small Language Models (SLMs) are emerging as viable alternatives for resource-constrained edge environments, though they often lack the capabilities of their larger counterparts. This article explores the potential of collaborative inference systems that leverage both edge and cloud resources to address these challenges. By presenting distinct cooperation strategies alongside practical design principles and experimental insights, we offer actionable guidance for deploying GenAI across the computing continuum.
The combination of Integrated Sensing and Communication (ISAC) and Mobile Edge Computing (MEC) enables devices to simultaneously sense the environment and offload data to the base stations (BS) for intelligent processing, thereby reducing local computational burdens. However, transmitting raw sensing data from ISAC devices to the BS often incurs substantial fronthaul overhead and latency. This paper investigates a three-tier collaborative inference framework enabled by Integrated Sensing, Communication, and Computing (ISCC), where cloud servers, MEC servers, and ISAC devices cooperatively execute different segments of a pre-trained deep neural network (DNN) for intelligent sensing. By offloading intermediate DNN features, the proposed framework can significantly reduce fronthaul transmission load. Furthermore, multiple-input multiple-output (MIMO) technology is employed to enhance both sensing quality and offloading efficiency. To minimize the overall sensing task inference latency across all ISAC devices, we jointly optimize the DNN partitioning strategy, ISAC beamforming, and computational resource allocation at the MEC servers and devices, subject to sensing beampattern constraints. We also propose an efficient two-layer optimization algorithm. In the inner layer, we derive closed-form solutions for computational resource allocation using the Karush-Kuhn-Tucker conditions. Moreover, we design the ISAC beamforming vectors via an iterative method based on the majorization-minimization and weighted minimum mean square error techniques. In the outer layer, we develop a cross-entropy based probabilistic learning algorithm to determine an optimal DNN partitioning strategy. Simulation results demonstrate that the proposed framework substantially outperforms existing two-tier schemes in inference latency.
Plant disease is a critical factor affecting agricultural production. Traditional manual recognition methods face significant drawbacks, including low accuracy, high costs, and inefficiency. Deep learning techniques have demonstrated significant benefits in identifying plant diseases, but they still face challenges such as inference delays and high energy consumption. Deep learning algorithms are difficult to run on resource-limited embedded devices. Offloading these models to cloud servers is confronted with the restriction of communication bandwidth, and all of these factors will influence the inference's efficiency. We propose a collaborative inference framework for recognizing plant diseases between edge devices and cloud servers to enhance inference speed. The DNN model for plant disease recognition is pruned through deep reinforcement learning to improve the inference speed and reduce energy consumption. Then the optimal split point is determined by a greedy strategy to achieve the best collaborated inference acceleration. Finally, the system for collaborative inference acceleration in plant disease recognition has been implemented using Gradio to facilitate friendly human-machine interaction. Experiments indicate that the proposed collaborative inference framework significantly increases inference speed while maintaining acceptable recognition accuracy, offering a novel solution for rapidly diagnosing and preventing plant diseases.
Generative large language models (LLMs) have garnered significant attention due to their exceptional capabilities in various AI tasks. Traditionally deployed in cloud datacenters, LLMs are now increasingly moving towards more accessible edge platforms to protect sensitive user data and ensure privacy preservation. The limited computational resources of individual edge devices, however, can result in excessively prolonged inference latency and overwhelmed memory usage. While existing research has explored collaborative edge computing to break the resource wall of individual devices, these solutions yet suffer from massive communication overhead and under-utilization of edge resources. Furthermore, they focus exclusively on optimizing the prefill phase, neglecting the crucial autoregressive decoding phase for generative LLMs. To address that, we propose Jupiter, a fast, scalable, and resource-efficient collaborative edge AI system for generative LLM inference. Jupiter introduces a flexible pipelined architecture as a principle and differentiates its system design according to the differentiated characteristics of the prefill and decoding phases. For prefill phase, Jupiter submits a novel intra-sequence pipeline parallelism and develops a meticulous parallelism planning strategy to maximize resource efficiency; For decoding, Jupiter devises an effective outline-based pipeline parallel decoding mechanism combined with speculative decoding, which further magnifies inference acceleration. Extensive evaluation based on realistic implementation demonstrates that Jupiter remarkably outperforms state-of-the-art approaches under various edge environment setups, achieving up to 26.1x end-to-end latency reduction while rendering on-par generation quality.
Spoken dialogue systems powered by large language models have demonstrated remarkable abilities in understanding human speech and generating appropriate spoken responses. However, these systems struggle with end-turn detection (ETD) -- the ability to distinguish between user turn completion and hesitation. This limitation often leads to premature or delayed responses, disrupting the flow of spoken conversations. In this paper, we introduce the ETD Dataset, the first public dataset for end-turn detection. The ETD dataset consists of both synthetic speech data generated with text-to-speech models and real-world speech data collected from web sources. We also propose SpeculativeETD, a novel collaborative inference framework that balances efficiency and accuracy to improve real-time ETD in resource-constrained environments. Our approach jointly employs a lightweight GRU-based model, which rapidly detects the non-speaking units in real-time on local devices, and a high-performance Wav2vec-based model running on the server to make a more challenging classification of distinguishing turn ends from mere pauses. Experiments demonstrate that the proposed SpeculativeETD significantly improves ETD accuracy while keeping the required computations low. Datasets and code will be available after the review.
Due to the limited computational resources, most Large Language Models (LLMs) developers can only fine-tune Small Language Models (SLMs) on their own data. These private SLMs typically have limited effectiveness. To boost the performance of private SLMs, this paper proposes to ask general LLMs for help. The general LLMs can be APIs or larger LLMs whose inference cost the developers can afford. Specifically, we propose the G-Boost framework where a private SLM adaptively performs collaborative inference with a general LLM under the guide of process reward. Experiments demonstrate that our framework can significantly boost the performance of private SLMs.
By locally encoding raw data into intermediate features, collaborative inference enables end users to leverage powerful deep learning models without exposure of sensitive raw data to cloud servers. However, recent studies have revealed that these intermediate features may not sufficiently preserve privacy, as information can be leaked and raw data can be reconstructed via model inversion attacks (MIAs). Obfuscation-based methods, such as noise corruption, adversarial representation learning, and information filters, enhance the inversion robustness by obfuscating the task-irrelevant redundancy empirically. However, methods for quantifying such redundancy remain elusive, and the explicit mathematical relation between this redundancy minimization and inversion robustness enhancement has not yet been established. To address that, this work first theoretically proves that the conditional entropy of inputs given intermediate features provides a guaranteed lower bound on the reconstruction mean square error (MSE) under any MIA. Then, we derive a differentiable and solvable measure for bounding this conditional entropy based on the Gaussian mixture estimation and propose a conditional entropy maximization (CEM) algorithm to enhance the inversion robustness. Experimental results on four datasets demonstrate the effectiveness and adaptability of our proposed CEM; without compromising feature utility and computing efficiency, plugging the proposed CEM into obfuscation-based defense mechanisms consistently boosts their inversion robustness, achieving average gains ranging from 12.9\% to 48.2\%. Code is available at \href{https://github.com/xiasong0501/CEM}{https://github.com/xiasong0501/CEM}.
Distributed DNN inference is becoming increasingly important as the demand for intelligent services at the network edge grows. By leveraging the power of distributed computing, edge devices can perform complicated and resource-hungry inference tasks previously only possible on powerful servers, enabling new applications in areas such as autonomous vehicles, industrial automation, and smart homes. However, it is challenging to achieve accurate and efficient distributed edge inference due to the fluctuating nature of the actual resources of the devices and the processing difficulty of the input data. In this work, we propose DistrEE, a distributed DNN inference framework that can exit model inference early to meet specific quality of service requirements. In particular, the framework firstly integrates model early exit and distributed inference for multi-node collaborative inferencing scenarios. Furthermore, it designs an early exit policy to control when the model inference terminates. Extensive simulation results demonstrate that DistrEE can efficiently realize efficient collaborative inference, achieving an effective trade-off between inference latency and accuracy.
Learning informative representations of phylogenetic tree structures is essential for analyzing evolutionary relationships. Classical distance-based methods have been widely used to project phylogenetic trees into Euclidean space, but they are often sensitive to the choice of distance metric and may lack sufficient resolution. In this paper, we introduce phylogenetic variational autoencoders (PhyloVAEs), an unsupervised learning framework designed for representation learning and generative modeling of tree topologies. Leveraging an efficient encoding mechanism inspired by autoregressive tree topology generation, we develop a deep latent-variable generative model that facilitates fast, parallelized topology generation. PhyloVAE combines this generative model with a collaborative inference model based on learnable topological features, allowing for high-resolution representations of phylogenetic tree samples. Extensive experiments demonstrate PhyloVAE's robust representation learning capabilities and fast generation of phylogenetic tree topologies.
Large language models have achieved remarkable success in various tasks but suffer from high computational costs during inference, limiting their deployment in resource-constrained applications. To address this issue, we propose a novel Collaborative Inference with Token-lEvel Routing (CITER) framework that enables efficient collaboration between small and large language models (SLMs \& LLMs) through a token-level routing strategy. Specifically, CITER routes non-critical tokens to an SLM for efficiency and routes critical tokens to an LLM for generalization quality. We formulate router training as a policy optimization, where the router receives rewards based on both the quality of predictions and the inference costs of generation. This allows the router to learn to predict token-level routing scores and make routing decisions based on both the current token and the future impact of its decisions. To further accelerate the reward evaluation process, we introduce a shortcut which significantly reduces the costs of the reward estimation and improving the practicality of our approach. Extensive experiments on five benchmark datasets demonstrate that CITER reduces the inference costs while preserving high-quality generation, offering a promising solution for real-time and resource-constrained applications. Our data and code are available at https://github.com/aiming-lab/CITER.
Large language models (LLMs) have demonstrated remarkable capabilities across a wide range of natural language processing tasks. Exploiting the heterogeneous capabilities of edge LLMs is crucial for diverse emerging applications, as it enables greater cost-effectiveness and reduced latency. In this work, we introduce \textit{Mixture-of-Edge-Experts (MoE$^2$)}, a novel collaborative inference framework for edge LLMs. We formulate the joint gating and expert selection problem to optimize inference performance under energy and latency constraints. Unlike conventional MoE problems, LLM expert selection is significantly more challenging due to the combinatorial nature and the heterogeneity of edge LLMs across various attributes. To this end, we propose a two-level expert selection mechanism through which we uncover an optimality-preserving property of gating parameters across expert selections. This property enables the decomposition of the training and selection processes, significantly reducing complexity. Furthermore, we leverage the objective's monotonicity and design a discrete monotonic optimization algorithm for optimal expert selection. We implement edge servers with NVIDIA Jetson AGX Orins and NVIDIA RTX 4090 GPUs, and perform extensive experiments. Our results validate that performance improvements of various LLM models and show that our MoE$^2$ method can achieve optimal trade-offs among different delay and energy budgets, and outperforms baselines under various system resource constraints.
Large Language Models (LLMs) for Recommendation (LLM4Rec) is a promising research direction that has demonstrated exceptional performance in this field. However, its inability to capture real-time user preferences greatly limits the practical application of LLM4Rec because (i) LLMs are costly to train and infer frequently, and (ii) LLMs struggle to access real-time data (its large number of parameters poses an obstacle to deployment on devices). Fortunately, small recommendation models (SRMs) can effectively supplement these shortcomings of LLM4Rec diagrams by consuming minimal resources for frequent training and inference, and by conveniently accessing real-time data on devices. In light of this, we designed the Device-Cloud LLM-SRM Collaborative Recommendation Framework (LSC4Rec) under a device-cloud collaboration setting. LSC4Rec aims to integrate the advantages of both LLMs and SRMs, as well as the benefits of cloud and edge computing, achieving a complementary synergy. We enhance the practicability of LSC4Rec by designing three strategies: collaborative training, collaborative inference, and intelligent request. During training, LLM generates candidate lists to enhance the ranking ability of SRM in collaborative scenarios and enables SRM to update adaptively to capture real-time user interests. During inference, LLM and SRM are deployed on the cloud and on the device, respectively. LLM generates candidate lists and initial ranking results based on user behavior, and SRM get reranking results based on the candidate list, with final results integrating both LLM's and SRM's scores. The device determines whether a new candidate list is needed by comparing the consistency of the LLM's and SRM's sorted lists. Our comprehensive and extensive experimental analysis validates the effectiveness of each strategy in LSC4Rec.
Mixture-of-Agents (MoA) has recently been proposed as a method to enhance performance of large language models (LLMs), enabling multiple individual LLMs to work together for collaborative inference. This collaborative approach results in improved responses to user prompts compared to relying on a single LLM. In this paper, we consider such an MoA architecture in a distributed setting, where LLMs operate on individual edge devices, each uniquely associated with a user and equipped with its own distributed computing power. These devices exchange information using decentralized gossip algorithms, allowing different device nodes to talk without the supervision of a centralized server. In the considered setup, different users have their own LLM models to address user prompts. Additionally, the devices gossip either their own user-specific prompts or augmented prompts to generate more refined answers to certain queries. User prompts are temporarily stored in the device queues when their corresponding LLMs are busy. Given the memory limitations of edge devices, it is crucial to ensure that the average queue sizes in the system remain bounded. In this paper, we address this by theoretically calculating the queuing stability conditions for the device queues under reasonable assumptions, which we validate experimentally as well. Further, we demonstrate through experiments, leveraging open-source LLMs for the implementation of distributed MoA, that certain MoA configurations produce higher-quality responses compared to others, as evaluated on AlpacaEval 2.0 benchmark. The implementation is available at: https://github.com/purbeshmitra/distributed_moa.
With the rapid advancement of artificial intelligence, generative artificial intelligence (GAI) has taken a leading role in transforming data processing methods. However, the high computational demands of GAI present challenges for devices with limited resources. As we move towards the sixth generation of mobile networks (6G), the higher data rates and improved energy efficiency of 6G create a need for more efficient data processing in GAI. Traditional GAI, however, shows its limitations in meeting these demands. To address these challenges, we introduce a multi-level collaborative inference system designed for next-generation networks and communication. Our proposed system features a deployment strategy that assigns models of varying sizes to devices at different network layers. Then, we design a task offloading strategy to optimise both efficiency and latency. Furthermore, a modified early exit mechanism is implemented to enhance the inference process for single models. Experimental results demonstrate that our system effectively reduces inference latency while maintaining high-quality output. Specifically, compared to existing work, our system can reduce inference time by up to 17% without sacrificing the inference accuracy.
Open-Vocabulary 3D object affordance grounding aims to anticipate ``action possibilities'' regions on 3D objects with arbitrary instructions, which is crucial for robots to generically perceive real scenarios and respond to operational changes. Existing methods focus on combining images or languages that depict interactions with 3D geometries to introduce external interaction priors. However, they are still vulnerable to a limited semantic space by failing to leverage implied invariant geometries and potential interaction intentions. Normally, humans address complex tasks through multi-step reasoning and respond to diverse situations by leveraging associative and analogical thinking. In light of this, we propose GREAT (GeometRy-intEntion collAboraTive inference) for Open-Vocabulary 3D Object Affordance Grounding, a novel framework that mines the object invariant geometry attributes and performs analogically reason in potential interaction scenarios to form affordance knowledge, fully combining the knowledge with both geometries and visual contents to ground 3D object affordance. Besides, we introduce the Point Image Affordance Dataset v2 (PIADv2), the largest 3D object affordance dataset at present to support the task. Extensive experiments demonstrate the effectiveness and superiority of GREAT. The code and dataset are available at https://yawen-shao.github.io/GREAT/.
Large Language Models (LLMs) exhibit remarkable human-like predictive capabilities. However, it is challenging to deploy LLMs to provide efficient and adaptive inference services at the edge. This paper proposes a novel Cloud-Edge Collaboration framework for LLMs (CE-CoLLM) to tackle these challenges. First, we identify the transmission of LLM contextual data between the cloud and edge as a key performance bottleneck, which introduces substantial communication overhead that dominates overall inference latency and makes na\"ive cloud-edge collaboration for LLMs inefficient. Second, we introduce a suite of novel techniques, including a latency-aware early exit mechanism and efficient cloud context management, into CE-CoLLM, which collectively reduce communication overhead and preserve LLM inference accuracy. Third, we design two adaptive inference modes to accommodate diverse edge environments: (1) a low-latency standalone edge inference mode that enables reliable edge-side independent LLM inference even under unstable network conditions, and (2) a high-accuracy cloud-edge collaborative inference mode that adaptively leverages cloud resources to enhance prediction accuracy. Extensive experiments on multiple benchmark datasets demonstrate that CE-CoLLM reduces overall inference time by up to 13.81% and offloads over 84.53% of the computational workload from the cloud to the edge, compared to conventional cloud-based LLM deployment, without sacrificing prediction accuracy. The code is provided on GitHub at https://github.com/mlsysx/CE-CoLLM.
Collaborative inference among multiple wireless edge devices has the potential to significantly enhance Artificial Intelligence (AI) applications, particularly for sensing and computer vision. This approach typically involves a three-stage process: a) data acquisition through sensing, b) feature extraction, and c) feature encoding for transmission. However, transmitting the extracted features poses a significant privacy risk, as sensitive personal data can be exposed during the process. To address this challenge, we propose a novel privacy-preserving collaborative inference mechanism, wherein each edge device in the network secures the privacy of extracted features before transmitting them to a central server for inference. Our approach is designed to achieve two primary objectives: 1) reducing communication overhead and 2) ensuring strict privacy guarantees during feature transmission, while maintaining effective inference performance. Additionally, we introduce an over-the-air pooling scheme specifically designed for classification tasks, which provides formal guarantees on the privacy of transmitted features and establishes a lower bound on classification accuracy.
Large language models (LLMs) have been a disruptive innovation in recent years, and they play a crucial role in our daily lives due to their ability to understand and generate human-like text. Their capabilities include natural language understanding, information retrieval and search, translation, chatbots, virtual assistance, and many more. However, it is well known that LLMs are massive in terms of the number of parameters. Additionally, the self-attention mechanism in the underlying architecture of LLMs, Transformers, has quadratic complexity in terms of both computation and memory with respect to the input sequence length. For these reasons, LLM inference is resource-intensive, and thus, the throughput of LLM inference is limited, especially for the longer sequences. In this report, we design a collaborative inference architecture between a server and its clients to alleviate the throughput limit. In this design, we consider the available resources on both sides, i.e., the computation and communication costs. We develop a dynamic programming-based algorithm to optimally allocate computation between the server and the client device to increase the server throughput, while not violating the service level agreement (SLA). We show in the experiments that we are able to efficiently distribute the workload allowing for roughly 1/3 reduction in the server workload, while achieving 19 percent improvement over a greedy method. As a result, we are able to demonstrate that, in an environment with different types of LLM inference requests, the throughput of the server is improved.
Motivated by the proliferation of Internet-of-Thing (IoT) devices and the rapid advances in the field of deep learning, there is a growing interest in pushing deep learning computations, conventionally handled by the cloud, to the edge of the network to deliver faster responses to end users, reduce bandwidth consumption to the cloud, and address privacy concerns. However, to fully realize deep learning at the edge, two main challenges still need to be addressed: (i) how to meet the high resource requirements of deep learning on resource-constrained devices, and (ii) how to leverage the availability of multiple streams of spatially correlated data, to increase the effectiveness of deep learning and improve application-level performance. To address the above challenges, we explore collaborative inference at the edge, in which edge nodes and end devices share correlated data and the inference computational burden by leveraging different ways to split computation and fuse data. Besides traditional centralized and distributed schemes for edge-end device collaborative inference, we introduce selective schemes that decrease bandwidth resource consumption by effectively reducing data redundancy. As a reference scenario, we focus on multi-view classification in a networked system in which sensing nodes can capture overlapping fields of view. The proposed schemes are compared in terms of accuracy, computational expenditure at the nodes, communication overhead, inference latency, robustness, and noise sensitivity. Experimental results highlight that selective collaborative schemes can achieve different trade-offs between the above performance metrics, with some of them bringing substantial communication savings (from 18% to 74% of the transmitted data with respect to centralized inference) while still keeping the inference accuracy well above 90%.