Open-source Large Language models (OsLLMs) propel the democratization of natural language research by giving the flexibility to augment or update model parameters for performance improvement. Nevertheless, like proprietary LLMs, Os-LLMs offer poorer performance on low-resource languages (LRLs) than high-resource languages (HRLs), owing to smaller amounts of training data and underrepresented vocabulary. On the other hand, continual pre-training (CPT) with large amounts of language-specific data is a costly proposition in terms of data acquisition and computational resources. Our goal is to drastically reduce CPT cost. To that end, we first develop a new algorithm to select a subset of texts from a larger corpus. We show the effectiveness of our technique using very little CPT data. In search of further improvement, we design a new algorithm to select tokens to include in the LLM vocabulary. We experiment with the recent Llama-3 model and nine Indian languages with diverse scripts and extent of resource availability. For evaluation, we use IndicGenBench, a generation task benchmark dataset for Indic languages. We experiment with various CPT corpora and augmented vocabulary size and offer insights across language families.

As large language models (LLMs) are shaping the way information is shared and accessed online, their opinions have the potential to influence a wide audience. This study examines who the LLMs view as the most prominent figures across various fields, using prompts in ten different languages to explore the influence of linguistic diversity. Our findings reveal low diversity in responses, with a small number of figures dominating recognition across languages (also known as the "superstar effect"). These results highlight the risk of narrowing global knowledge representation when LLMs retrieve subjective information.

Low bit-precisions and their bit-slice sparsity have recently been studied to accelerate general matrix-multiplications (GEMM) during large-scale deep neural network (DNN) inferences. While the conventional symmetric quantization facilitates low-resolution processing with bit-slice sparsity for both weight and activation, its accuracy loss caused by the activation's asymmetric distributions cannot be acceptable, especially for large-scale DNNs. In efforts to mitigate this accuracy loss, recent studies have actively utilized asymmetric quantization for activations without requiring additional operations. However, the cutting-edge asymmetric quantization produces numerous nonzero slices that cannot be compressed and skipped by recent bit-slice GEMM accelerators, naturally consuming more processing energy to handle the quantized DNN models. To simultaneously achieve high accuracy and hardware efficiency for large-scale DNN inferences, this paper proposes an Asymmetrically-Quantized bit-Slice GEMM (AQS-GEMM) for the first time. In contrast to the previous bit-slice computing, which only skips operations of zero slices, the AQS-GEMM compresses frequent nonzero slices, generated by asymmetric quantization, and skips their operations. To increase the slice-level sparsity of activations, we also introduce two algorithm-hardware co-optimization methods: a zero-point manipulation and a distribution-based bit-slicing. To support the proposed AQS-GEMM and optimizations at the hardware-level, we newly introduce a DNN accelerator, Panacea, which efficiently handles sparse/dense workloads of the tiled AQS-GEMM to increase data reuse and utilization. Panacea supports a specialized dataflow and run-length encoding to maximize data reuse and minimize external memory accesses, significantly improving its hardware efficiency. Our benchmark evaluations show Panacea outperforms existing DNN accelerators.

RAG (Retrieval Augmented Generation) allows LLMs (large language models) to generate better responses with external knowledge, but using more external knowledge often improves generation quality at the expense of response delay. Prior work either reduces the response delay (through better scheduling of RAG queries) or strives to maximize quality (which involves tuning the RAG workflow), but they fall short in optimizing the tradeoff between the delay and quality of RAG responses. This paper presents RAGServe, the first RAG system that jointly schedules queries and adapts the key RAG configurations of each query, such as the number of retrieved text chunks and synthesis methods, in order to balance quality optimization and response delay reduction. Using 4 popular RAG-QA datasets, we show that compared with the state-of-the-art RAG optimization schemes, RAGServe reduces the generation latency by $1.64-2.54\times$ without sacrificing generation quality.

Large Multimodal Models (LMMs) have demonstrated exceptional comprehension and interpretation capabilities in Autonomous Driving (AD) by incorporating large language models. Despite the advancements, current data-driven AD approaches tend to concentrate on a single dataset and specific tasks, neglecting their overall capabilities and ability to generalize. To bridge these gaps, we propose DriveMM, a general large multimodal model designed to process diverse data inputs, such as images and multi-view videos, while performing a broad spectrum of AD tasks, including perception, prediction, and planning. Initially, the model undergoes curriculum pre-training to process varied visual signals and perform basic visual comprehension and perception tasks. Subsequently, we augment and standardize various AD-related datasets to fine-tune the model, resulting in an all-in-one LMM for autonomous driving. To assess the general capabilities and generalization ability, we conduct evaluations on six public benchmarks and undertake zero-shot transfer on an unseen dataset, where DriveMM achieves state-of-the-art performance across all tasks. We hope DriveMM as a promising solution for future end-to-end autonomous driving applications in the real world. Project page with code: https://github.com/zhijian11/DriveMM.

Large Language Models (LLMs) have become the new foundation for many applications, reshaping human society like a storm. Disaggregated inference, which separates prefill and decode stages, is a promising approach to improving hardware utilization and service quality. However, due to inefficient inter-node communication, existing systems restrict disaggregated inference to a single node, limiting resource allocation flexibility and reducing service capacity. This paper introduces KVDirect, which optimizes KV cache transfer to enable a distributed disaggregated LLM inference. KVDirect achieves this through the following contributions. First, we propose a novel tensor-centric communication mechanism that reduces the synchronization overhead in traditional distributed GPU systems. Second, we design a custom communication library to support dynamic GPU resource scheduling and efficient KV cache transfer. Third, we introduce a pull-based KV cache transfer strategy that reduces GPU resource idling and improves latency. Finally, we implement KVDirect as an open-source LLM inference framework. Our evaluation demonstrates that KVDirect reduces per-request latency by 55% compared to the baseline across diverse workloads under the same resource constraints.

Large Language Models (LLMs) are transforming mental health care by enhancing accessibility, personalization, and efficiency in therapeutic interventions. These AI-driven tools empower mental health professionals with real-time support, improved data integration, and the ability to encourage care-seeking behaviors, particularly in underserved communities. By harnessing LLMs, practitioners can deliver more empathetic, tailored, and effective support, addressing longstanding gaps in mental health service provision. However, their implementation comes with significant challenges and ethical concerns. Performance limitations, data privacy risks, biased outputs, and the potential for generating misleading information underscore the critical need for stringent ethical guidelines and robust evaluation mechanisms. The sensitive nature of mental health data further necessitates meticulous safeguards to protect patient rights and ensure equitable access to AI-driven care. Proponents argue that LLMs have the potential to democratize mental health resources, while critics warn of risks such as misuse and the diminishment of human connection in therapy. Achieving a balance between innovation and ethical responsibility is imperative. This paper examines the transformative potential of LLMs in mental health care, highlights the associated technical and ethical complexities, and advocates for a collaborative, multidisciplinary approach to ensure these advancements align with the goal of providing compassionate, equitable, and effective mental health support.

Since decades, the modelling of metadata has been core to the functioning of any academic library. Its importance has only enhanced with the increasing pervasiveness of Generative Artificial Intelligence (AI)-driven information activities and services which constitute a library's outreach. However, with the rising importance of metadata, there arose several outstanding problems with the process of designing a library metadata model impacting its reusability, crosswalk and interoperability with other metadata models. This paper posits that the above problems stem from an underlying thesis that there should only be a few core metadata models which would be necessary and sufficient for any information service using them, irrespective of the heterogeneity of intra-domain or inter-domain settings. To that end, this paper advances a contrary view of the above thesis and substantiates its argument in three key steps. First, it introduces a novel way of thinking about a library metadata model as an ontology-driven composition of five functionally interlinked representation levels from perception to its intensional definition via properties. Second, it introduces the representational manifoldness implicit in each of the five levels which cumulatively contributes to a conceptually entangled library metadata model. Finally, and most importantly, it proposes a Generative AI-driven Human-Large Language Model (LLM) collaboration based metadata modelling approach to disentangle the entanglement inherent in each representation level leading to the generation of a conceptually disentangled metadata model. Throughout the paper, the arguments are exemplified by motivating scenarios and examples from representative libraries handling cancer information.

The capability of Large Language Models (LLMs) to plan remains a topic of debate. Some critics argue that strategies to boost LLMs' reasoning skills are ineffective in planning tasks, while others report strong outcomes merely from training models on a planning corpus. This study reassesses recent strategies by developing an end-to-end LLM planner and employing diverse metrics for a thorough evaluation. We find that merely fine-tuning LLMs on a corpus of planning instances does not lead to robust planning skills, as indicated by poor performance on out-of-distribution test sets. At the same time, we find that various strategies, including Chain-of-Thought, do enhance the probability of a plan being executable. This indicates progress towards better plan quality, despite not directly enhancing the final validity rate. Among the strategies we evaluated, reinforcement learning with our novel `Longest Contiguous Common Subsequence' reward emerged as the most effective, contributing to both plan validity and executability. Overall, our research addresses key misconceptions in the LLM-planning literature; we validate incremental progress in plan executability, although plan validity remains a challenge. Hence, future strategies should focus on both these aspects, drawing insights from our findings.

Digital agents are increasingly employed to automate tasks in interactive digital environments such as web pages, software applications, and operating systems. While text-based agents built on Large Language Models (LLMs) often require frequent updates due to platform-specific APIs, visual agents leveraging Multimodal Large Language Models (MLLMs) offer enhanced adaptability by interacting directly with Graphical User Interfaces (GUIs). However, these agents face significant challenges in visual perception, particularly when handling high-resolution, visually complex digital environments. This paper introduces Iris, a foundational visual agent that addresses these challenges through two key innovations: Information-Sensitive Cropping (ISC) and Self-Refining Dual Learning (SRDL). ISC dynamically identifies and prioritizes visually dense regions using a edge detection algorithm, enabling efficient processing by allocating more computational resources to areas with higher information density. SRDL enhances the agent's ability to handle complex tasks by leveraging a dual-learning loop, where improvements in referring (describing UI elements) reinforce grounding (locating elements) and vice versa, all without requiring additional annotated data. Empirical evaluations demonstrate that Iris achieves state-of-the-art performance across multiple benchmarks with only 850K GUI annotations, outperforming methods using 10x more training data. These improvements further translate to significant gains in both web and OS agent downstream tasks.