RAG systems rely on rerankers to identify relevant documents. However, fine-tuning these models remains challenging due to the scarcity of annotated query-document pairs. Existing distillation-based approaches suffer from training-inference misalignment and fail to capture interdependencies among candidate documents. To overcome these limitations, we reframe the reranking process as an attention-mask problem and propose Gumbel Reranking, an end-to-end training framework for rerankers aimed at minimizing the training-inference gap. In our approach, reranker optimization is reformulated as learning a stochastic, document-wise Top-$k$ attention mask using the Gumbel Trick and Relaxed Top-$k$ Sampling. This formulation enables end-to-end optimization by minimizing the overall language loss. Experiments across various settings consistently demonstrate performance gains, including a 10.4\% improvement in recall on HotpotQA for distinguishing indirectly relevant documents.

A wide range of user perception applications leverage inertial measurement unit (IMU) data for online prediction. However, restricted by the non-i.i.d. nature of IMU data collected from mobile devices, most systems work well only in a controlled setting (e.g., for a specific user in particular postures), limiting application scenarios. To achieve uncontrolled online prediction on mobile devices, referred to as the flexible user perception (FUP) problem, is attractive but hard. In this paper, we propose a novel scheme, called Prism, which can obtain high FUP accuracy on mobile devices. The core of Prism is to discover task-aware domains embedded in IMU dataset, and to train a domain-aware model on each identified domain. To this end, we design an expectation-maximization (EM) algorithm to estimate latent domains with respect to the specific downstream perception task. Finally, the best-fit model can be automatically selected for use by comparing the test sample and all identified domains in the feature space. We implement Prism on various mobile devices and conduct extensive experiments. Results demonstrate that Prism can achieve the best FUP performance with a low latency.

The emergence of large-scale Mixture of Experts (MoE) models has marked a significant advancement in artificial intelligence, offering enhanced model capacity and computational efficiency through conditional computation. However, the deployment and inference of these models present substantial challenges in terms of computational resources, latency, and energy efficiency. This comprehensive survey systematically analyzes the current landscape of inference optimization techniques for MoE models across the entire system stack. We first establish a taxonomical framework that categorizes optimization approaches into model-level, system-level, and hardware-level optimizations. At the model level, we examine architectural innovations including efficient expert design, attention mechanisms, various compression techniques such as pruning, quantization, and knowledge distillation, as well as algorithm improvement including dynamic routing strategies and expert merging methods. At the system level, we investigate distributed computing approaches, load balancing mechanisms, and efficient scheduling algorithms that enable scalable deployment. Furthermore, we delve into hardware-specific optimizations and co-design strategies that maximize throughput and energy efficiency. This survey not only provides a structured overview of existing solutions but also identifies key challenges and promising research directions in MoE inference optimization. Our comprehensive analysis serves as a valuable resource for researchers and practitioners working on large-scale deployment of MoE models in resource-constrained environments. To facilitate ongoing updates and the sharing of cutting-edge advances in MoE inference optimization research, we have established a repository accessible at \url{https://github.com/MoE-Inf/awesome-moe-inference/}.

Large Language Models (LLMs) have been widely deployed in a variety of applications, and the context length is rapidly increasing to handle tasks such as long-document QA and complex logical reasoning. However, long context poses significant challenges for inference efficiency, including high memory costs of key-value (KV) cache and increased latency due to extensive memory accesses. Recent works have proposed compressing KV cache to approximate computation, but these methods either evict tokens permanently, never recalling them for later inference, or recall previous tokens at the granularity of pages divided by textual positions. Both approaches degrade the model accuracy and output quality. To achieve efficient and accurate recallable KV cache compression, we introduce ClusterKV, which recalls tokens at the granularity of semantic clusters. We design and implement efficient algorithms and systems for clustering, selection, indexing and caching. Experiment results show that ClusterKV attains negligible accuracy loss across various tasks with 32k context lengths, using only a 1k to 2k KV cache budget, and achieves up to a 2$\times$ speedup in latency and a 2.5$\times$ improvement in decoding throughput. Compared to SoTA recallable KV compression methods, ClusterKV demonstrates higher model accuracy and output quality, while maintaining or exceeding inference efficiency.

Transformer models have gained significant attention due to their power in machine learning tasks. Their extensive deployment has raised concerns about the potential leakage of sensitive information during inference. However, when being applied to Transformers, existing approaches based on secure two-party computation (2PC) bring about efficiency limitations in two folds: (1) resource-intensive matrix multiplications in linear layers, and (2) complex non-linear activation functions like $\mathsf{GELU}$ and $\mathsf{Softmax}$. This work presents a new two-party inference framework $\mathsf{Nimbus}$ for Transformer models. For the linear layer, we propose a new 2PC paradigm along with an encoding approach to securely compute matrix multiplications based on an outer-product insight, which achieves $2.9\times \sim 12.5\times$ performance improvements compared to the state-of-the-art (SOTA) protocol. For the non-linear layer, through a new observation of utilizing the input distribution, we propose an approach of low-degree polynomial approximation for $\mathsf{GELU}$ and $\mathsf{Softmax}$, which improves the performance of the SOTA polynomial approximation by $2.9\times \sim 4.0\times$, where the average accuracy loss of our approach is 0.08\% compared to the non-2PC inference without privacy. Compared with the SOTA two-party inference, $\mathsf{Nimbus}$ improves the end-to-end performance of \bert{} inference by $2.7\times \sim 4.7\times$ across different network settings.

The Mixture-of-Experts (MoE) architecture has demonstrated significant advantages in the era of Large Language Models (LLMs), offering enhanced capabilities with reduced inference costs. However, deploying MoE-based LLMs on memoryconstrained edge devices remains challenging due to their substantial memory requirements. While existing expertoffloading methods alleviate the memory requirements, they often incur significant expert-loading costs or compromise model accuracy. We present HOBBIT, a mixed precision expert offloading system to enable flexible and efficient MoE inference. Our key insight is that dynamically replacing less critical cache-miss experts with low precision versions can substantially reduce expert-loading latency while preserving model accuracy. HOBBIT introduces three innovative techniques that map the natural hierarchy of MoE computation: (1) a token-level dynamic expert loading mechanism, (2) a layer-level adaptive expert prefetching technique, and (3) a sequence-level multidimensional expert caching policy. These innovations fully leverage the benefits of mixedprecision expert inference. By implementing HOBBIT on top of the renowned LLM inference framework Llama.cpp, we evaluate its performance across different edge devices with representative MoE models. The results demonstrate that HOBBIT achieves up to a 9.93x speedup in decoding compared to state-of-the-art MoE offloading systems.

Emerging Artificial Intelligence of Things (AIoT) applications desire online prediction using deep neural network (DNN) models on mobile devices. However, due to the movement of devices, unfamiliar test samples constantly appear, significantly affecting the prediction accuracy of a pre-trained DNN. In addition, unstable network connection calls for local model inference. In this paper, we propose a light-weight scheme, called Anole, to cope with the local DNN model inference on mobile devices. The core idea of Anole is to first establish an army of compact DNN models, and then adaptively select the model fitting the current test sample best for online inference. The key is to automatically identify model-friendly scenes for training scene-specific DNN models. To this end, we design a weakly-supervised scene representation learning algorithm by combining both human heuristics and feature similarity in separating scenes. Moreover, we further train a model classifier to predict the best-fit scene-specific DNN model for each test sample. We implement Anole on different types of mobile devices and conduct extensive trace-driven and real-world experiments based on unmanned aerial vehicles (UAVs). The results demonstrate that Anole outwits the method of using a versatile large DNN in terms of prediction accuracy (4.5% higher), response time (33.1% faster) and power consumption (45.1% lower).

Joint consideration of scheduling and adaptive parallelism offers great opportunities for improving the training efficiency of large models on heterogeneous GPU clusters. However, integrating adaptive parallelism into a cluster scheduler expands the cluster scheduling space. The new space is the product of the original scheduling space and the parallelism exploration space of adaptive parallelism (also a product of pipeline, data, and tensor parallelism). The exponentially enlarged scheduling space and ever-changing optimal parallelism plan from adaptive parallelism together result in the contradiction between low-overhead and accurate performance data acquisition for efficient cluster scheduling. This paper presents Crius, a training system for efficiently scheduling multiple large models with adaptive parallelism in a heterogeneous cluster. Crius proposes a novel scheduling granularity called Cell. It represents a job with deterministic resources and pipeline stages. The exploration space of Cell is shrunk to the product of only data and tensor parallelism, thus exposing the potential for accurate and low-overhead performance estimation. Crius then accurately estimates Cells and efficiently schedules training jobs. When a Cell is selected as a scheduling choice, its represented job runs with the optimal parallelism plan explored. Experimental results show that Crius reduces job completion time by up to 48.9% and schedules large models with up to 1.49x cluster throughput improvement.

High-level synthesis (HLS) notably speeds up the hardware design process by avoiding RTL programming. However, the turnaround time of HLS increases significantly when post-route quality of results (QoR) are considered during optimization. To tackle this issue, we propose a hierarchical post-route QoR prediction approach for FPGA HLS, which features: (1) a modeling flow that directly estimates latency and post-route resource usage from C/C++ programs; (2) a graph construction method that effectively represents the control and data flow graph of source code and effects of HLS pragmas; and (3) a hierarchical GNN training and prediction method capable of capturing the impact of loop hierarchies. Experimental results show that our method presents a prediction error of less than 10% for different types of QoR metrics, which gains tremendous improvement compared with the state-of-the-art GNN methods. By adopting our proposed methodology, the runtime for design space exploration in HLS is shortened to tens of minutes and the achieved ADRS is reduced to 6.91% on average.

Post-training quantization attracts increasing attention due to its convenience in deploying quantized neural networks. Although rounding-to-nearest remains the prevailing method for DNN quantization, prior research has demonstrated its suboptimal nature when applied to weight quantization. They propose optimizing weight rounding schemes by leveraging output error rather than the traditional weight quantization error. Our study reveals that similar rounding challenges also extend to activation quantization. Despite the easy generalization, the challenges lie in the dynamic nature of activation. Adaptive rounding is expected for varying activations and the method is subjected to runtime overhead. To tackle this, we propose the AQuant quantization framework with a novel perspective to reduce output error by adjusting rounding schemes of activations. Instead of using the constant rounding border 0.5 of the rounding-to-nearest operation, we make the border become a function w.r.t. the activation value to change the activation rounding by the adaptive border. To deal with the runtime overhead, we use a coarse-grained version of the border function. Finally, we introduce our framework to optimize the border function. Extensive experiments show that AQuant achieves notable improvements compared to state-of-the-art works and pushes the accuracy of ResNet-18 up to 60.31% under the 2-bit weight and activation quantization.