To support latency-sensitive AI applications ranging from autonomous driving to industrial robot manipulation, 6G envisions distributed ML with computational resources in mobile, edge, and cloud connected over hyper-reliable low-latency communication (HRLLC). In this setting, speculative decoding can facilitate collaborative inference of models distributively deployed: a lightweight on-device model locally generates drafts while a more capable remote target model on a server verifies and corrects them in parallel with speculative sampling, thus resulting in lower latency without compromising accuracy. However, unlike autoregressive text generation, behavior cloning policies, typically used for embodied AI applications, cannot parallelize verification and correction for multiple drafts as each generated action depends on observation updated by a previous action. To this end, we propose Action Deviation-Aware Hybrid Inference (ADAHI), wherein drafts are selectively transmitted and verified based on action deviation, which has a strong correlation with action's rejection probability by the target model. By invoking server operation only when necessary, communication and computational overhead can be reduced while accuracy gain from speculative sampling is preserved. Experiments on our testbed show that ADAHI reduces transmission and server operations by approximately 40%, lowers end-to-end latency by 39.2%, and attains up to 97.2% of the task-success rate of baseline that invokes speculative sampling for every draft embedding vector.

Diffusion models (DMs) have emerged as powerful tools for high-quality content generation, yet their intensive computational requirements for inference pose challenges for resource-constrained edge devices. Cloud-based solutions aid in computation but often fall short in addressing privacy risks, personalization efficiency, and communication costs in multi-user edge-AIGC scenarios. To bridge this gap, we first analyze existing edge-AIGC applications in personalized content synthesis, revealing their limitations in efficiency and scalability. We then propose a novel cluster-aware hierarchical federated aggregation framework. Based on parameter-efficient local fine-tuning via Low-Rank Adaptation (LoRA), the framework first clusters clients based on the similarity of their uploaded task requirements, followed by an intra-cluster aggregation for enhanced personalization at the server-side. Subsequently, an inter-cluster knowledge interaction paradigm is implemented to enable hybrid-style content generation across diverse clusters.Building upon federated learning (FL) collaboration, our framework simultaneously trains personalized models for individual users at the devices and a shared global model enhanced with multiple LoRA adapters on the server,enabling efficient edge inference; meanwhile, all prompts for clustering and inference are encoded prior to transmission, thereby further mitigating the risk of plaintext leakage. Our evaluations demonstrate that the framework achieves accelerated convergence while maintaining practical viability for scalable multi-user personalized AIGC services under edge constraints.

Overall this is a nice idea that works on using black box models to amortize the residuals from doing inference assuming a linearized approximation to the model. I found the experiments to be well organized albeit mostly on small scale/synthetic data. Summary: This paper introduces a procedure for combining graph neural networks with traditional methods for probabilistic inference (instantiated in HMMs). When we have linear dynamics in a HMM, inference is exact. For nonlinear dynamics, when we have access to the functional form of the true dynamics of the state space model, we can linearize the transition and emission functions (via a Taylor expansion) and represent them as matrices.

A graphical model is a structured representation of the data generating process. The traditional method to reason over random variables is to perform inference in this graphical model. However, in many cases the generating process is only a poor approximation of the much more complex true data generating process, leading to suboptimal estimation. The subtleties of the generative process are however captured in the data itself and we can learn to infer'', that is, learn a direct mapping from observations to explanatory latent variables. In this work we propose a hybrid model that combines graphical inference with a learned inverse model, which we structure as in a graph neural network, while the iterative algorithm as a whole is formulated as a recurrent neural network. By using cross-validation we can automatically balance the amount of work performed by graphical inference versus learned inference.

The customization of large language models (LLMs) for user-specified tasks gets important. However, maintaining all the customized LLMs on cloud servers incurs substantial memory and computational overheads, and uploading user data can also lead to privacy concerns. On-device LLMs can offer a promising solution by mitigating these issues. Yet, the performance of on-device LLMs is inherently constrained by the limitations of small-scaled models. To overcome these restrictions, we first propose Crayon, a novel approach for on-device LLM customization. Crayon begins by constructing a pool of diverse base adapters, and then we instantly blend them into a customized adapter without extra training. In addition, we develop a device-server hybrid inference strategy, which deftly allocates more demanding queries or non-customized tasks to a larger, more capable LLM on a server. This ensures optimal performance without sacrificing the benefits of on-device customization. We carefully craft a novel benchmark from multiple question-answer datasets, and show the efficacy of our method in the LLM customization.

A graphical model is a structured representation of the data generating process. The traditional method to reason over random variables is to perform inference in this graphical model. However, in many cases the generating process is only a poor approximation of the much more complex true data generating process, leading to suboptimal estimation. The subtleties of the generative process are however captured in the data itself and we can learn to infer'', that is, learn a direct mapping from observations to explanatory latent variables. In this work we propose a hybrid model that combines graphical inference with a learned inverse model, which we structure as in a graph neural network, while the iterative algorithm as a whole is formulated as a recurrent neural network.