Retrieval-augmented generation (RAG) extends large language models (LLMs) with external data sources to enhance factual correctness and domain coverage. Modern RAG pipelines rely on large datastores, leading to system challenges in latency-sensitive deployments, especially when limited GPU memory is available. To address these challenges, we propose TeleRAG, an efficient inference system that reduces RAG latency with minimal GPU memory requirements. The core innovation of TeleRAG is lookahead retrieval, a prefetching mechanism that anticipates required data and transfers it from CPU to GPU in parallel with LLM generation. By leveraging the modularity of RAG pipelines, the inverted file index (IVF) search algorithm and similarities between queries, TeleRAG optimally overlaps data movement and computation. Experimental results show that TeleRAG reduces end-to-end RAG inference latency by up to 1.72x on average compared to state-of-the-art systems, enabling faster, more memory-efficient deployments of advanced RAG applications.

Long-context models are essential for many applications but face inefficiencies in loading large KV caches during decoding. Prior methods enforce fixed token budgets for sparse attention, assuming a set number of tokens can approximate full attention. However, these methods overlook variations in the importance of attention across heads, layers, and contexts. To address these limitations, we propose Tactic, a sparsity-adaptive and calibration-free sparse attention mechanism that dynamically selects tokens based on their cumulative attention scores rather than a fixed token budget. By setting a target fraction of total attention scores, Tactic ensures that token selection naturally adapts to variations in attention sparsity. To efficiently approximate this selection, Tactic leverages clustering-based sorting and distribution fitting, allowing it to accurately estimate token importance with minimal computational overhead. We show that Tactic outperforms existing sparse attention algorithms, achieving superior accuracy and up to 7.29x decode attention speedup. This improvement translates to an overall 1.58x end-to-end inference speedup, making Tactic a practical and effective solution for long-context LLM inference in accuracy-sensitive applications.

Observability in cloud infrastructure is critical for service providers, driving the widespread adoption of anomaly detection systems for monitoring metrics. However, existing systems often struggle to simultaneously achieve explainability, reproducibility, and autonomy, which are three indispensable properties for production use. We introduce Argos, an agentic system for detecting time-series anomalies in cloud infrastructure by leveraging large language models (LLMs). Argos proposes to use explainable and reproducible anomaly rules as intermediate representation and employs LLMs to autonomously generate such rules. The system will efficiently train error-free and accuracy-guaranteed anomaly rules through multiple collaborative agents and deploy the trained rules for low-cost online anomaly detection. Through evaluation results, we demonstrate that Argos outperforms state-of-the-art methods, increasing $F_1$ scores by up to $9.5\%$ and $28.3\%$ on public anomaly detection datasets and an internal dataset collected from Microsoft, respectively.

While textless Spoken Language Models (SLMs) have shown potential in end-to-end speech-to-speech modeling, they still lag behind text-based Large Language Models (LLMs) in terms of semantic coherence and relevance. This work introduces the Align-SLM framework, which leverages preference optimization inspired by Reinforcement Learning with AI Feedback (RLAIF) to enhance the semantic understanding of SLMs. Our approach generates multiple speech continuations from a given prompt and uses semantic metrics to create preference data for Direct Preference Optimization (DPO). We evaluate the framework using ZeroSpeech 2021 benchmarks for lexical and syntactic modeling, the spoken version of the StoryCloze dataset for semantic coherence, and other speech generation metrics, including the GPT4-o score and human evaluation. Experimental results show that our method achieves state-of-the-art performance for SLMs on most benchmarks, highlighting the importance of preference optimization to improve the semantics of SLMs.

Large Language Models (LLMs) based on Mixture-of-Experts (MoE) architecture are showing promising performance on various tasks. However, running them on resource-constrained settings, where GPU memory resources are not abundant, is challenging due to huge model sizes. Existing systems that offload model weights to CPU memory suffer from the significant overhead of frequently moving data between CPU and GPU. In this paper, we propose Fiddler, a resource-efficient inference engine with CPU-GPU orchestration for MoE models. The key idea of Fiddler is to use the computation ability of the CPU to minimize the data movement between the CPU and GPU. Our evaluation shows that Fiddler can run the uncompressed Mixtral-8x7B model, which exceeds 90GB in parameters, to generate over $3$ tokens per second on a single GPU with 24GB memory, showing an order of magnitude improvement over existing methods. The code of Fiddler is publicly available at \url{https://github.com/efeslab/fiddler}

Large Language Models (LLMs) have demonstrated superior abilities in tasks such as chatting, reasoning, and question-answering. However, standard LLMs may ignore crucial paralinguistic information, such as sentiment, emotion, and speaking style, which are essential for achieving natural, human-like spoken conversation, especially when such information is conveyed by acoustic cues. We therefore propose Paralinguistics-enhanced Generative Pretrained Transformer (ParalinGPT), an LLM that utilizes text and speech modalities to better model the linguistic content and paralinguistic attributes of spoken dialogue. The model takes the conversational context of text, speech embeddings, and paralinguistic attributes as input prompts within a serialized multitasking multimodal framework. Specifically, our framework serializes tasks in the order of current paralinguistic attribute prediction, response paralinguistic attribute prediction, and response text generation with autoregressive conditioning. We utilize the Switchboard-1 corpus, including its sentiment labels as the paralinguistic attribute, as our spoken dialogue dataset. Experimental results indicate the proposed serialized multitasking method outperforms typical sequence classification techniques on current and response sentiment classification. Furthermore, leveraging conversational context and speech embeddings significantly improves both response text generation and sentiment prediction. Our proposed framework achieves relative improvements of 6.7%, 12.0%, and 3.5% in current sentiment accuracy, response sentiment accuracy, and response text BLEU score, respectively.

In the realm of spoken language understanding (SLU), numerous natural language understanding (NLU) methodologies have been adapted by supplying large language models (LLMs) with transcribed speech instead of conventional written text. In real-world scenarios, prior to input into an LLM, an automated speech recognition (ASR) system generates an output transcript hypothesis, where inherent errors can degrade subsequent SLU tasks. Here we introduce a method that utilizes the ASR system's lattice output instead of relying solely on the top hypothesis, aiming to encapsulate speech ambiguities and enhance SLU outcomes. Our in-context learning experiments, covering spoken question answering and intent classification, underline the LLM's resilience to noisy speech transcripts with the help of word confusion networks from lattices, bridging the SLU performance gap between using the top ASR hypothesis and an oracle upper bound. Additionally, we delve into the LLM's robustness to varying ASR performance conditions and scrutinize the aspects of in-context learning which prove the most influential.

Training large AI models on numerous GPUs consumes a massive amount of energy. We observe that not all energy consumed during training directly contributes to end-to-end training throughput, and a significant portion can be removed without slowing down training, which we call energy bloat. In this work, we identify two independent sources of energy bloat in large model training, intrinsic and extrinsic, and propose Perseus, a unified optimization framework that mitigates both. Perseus obtains the "iteration time-energy" Pareto frontier of any large model training job using an efficient iterative graph cut-based algorithm and schedules energy consumption of its forward and backward computations across time to remove intrinsic and extrinsic energy bloat. Evaluation on large models like GPT-3 and Bloom shows that Perseus reduces energy consumption of large model training by up to 30%, enabling savings otherwise unobtainable before.

However, as the size of the pretrained models increases, the cost associated We propose a neural language modeling system based on with fine-tuning and deploying these models for low-rank adaptation (LoRA) for speech recognition output real-world applications also escalates. To address this practical rescoring. Although pretrained language models (LMs) challenge, a range of parameter-efficient methods (e.g., like BERT have shown superior performance in second-pass adapters, model reprogramming, and prompts) have been proposed rescoring, the high computational cost of scaling up the pretraining [11, 12, 13, 14, 15, 16, 17, 18] to alleviate the computation stage and adapting the pretrained models to specific and memory demands of fine-tuning LLMs. Low-rank domains limit their practical use in rescoring. Here we present adaptation (LoRA) [19] freezes all pretrained parameters in a method based on low-rank decomposition to train a rescoring the LLM and inserts a trainable pair of matrices (acting as a BERT model and adapt it to new domains using only a low-rank decomposition of a full matrix) additively into each fraction (0.08%) of the pretrained parameters. These inserted layer of the Transformer architecture. Compared to other matrices are optimized through a discriminative training objective parameter-efficient training methods, such as adapters [12], along with a correlation-based regularization loss. The LoRA has two distinct advantages: 1) it employs a simple proposed low-rank adaptation RescoreBERT (LoRB) architecture architecture and has the potential to reduce the number of is evaluated on LibriSpeech and internal datasets with trainable parameters compared to alternatives; 2) LoRA does decreased training times by factors between 5.4 and 3.6.

We explore the ability of large language models (LLMs) to act as speech recognition post-processors that perform rescoring and error correction. Our first focus is on instruction prompting to let LLMs perform these task without fine-tuning, for which we evaluate different prompting schemes, both zero- and few-shot in-context learning, and a novel task activation prompting method that combines causal instructions and demonstration to increase its context windows. Next, we show that rescoring only by in-context learning with frozen LLMs achieves results that are competitive with rescoring by domain-tuned LMs, using a pretrained first-pass recognition system and rescoring output on two out-of-domain tasks (ATIS and WSJ). By combining prompting techniques with fine-tuning we achieve error rates below the N-best oracle level, showcasing the generalization power of the LLMs.