While decoder-only large language models (LLMs) have shown impressive results, encoder-decoder models are still widely adopted in real-world applications for their inference efficiency and richer encoder representation. In this paper, we study a novel problem: adapting pretrained decoder-only LLMs to encoder-decoder, with the goal of leveraging the strengths of both approaches to achieve a more favorable quality-efficiency trade-off. We argue that adaptation not only enables inheriting the capability of decoder-only LLMs but also reduces the demand for computation compared to pretraining from scratch. We rigorously explore different pretraining objectives and parameter initialization/optimization techniques. Through extensive experiments based on Gemma 2 (2B and 9B) and a suite of newly pretrained mT5-sized models (up to 1.6B), we demonstrate the effectiveness of adaptation and the advantage of encoder-decoder LLMs. Under similar inference budget, encoder-decoder LLMs achieve comparable (often better) pretraining performance but substantially better finetuning performance than their decoder-only counterpart. For example, Gemma 2B-2B outperforms Gemma 2B by $\sim$7\% after instruction tuning. Encoder-decoder adaptation also allows for flexible combination of different-sized models, where Gemma 9B-2B significantly surpasses Gemma 2B-2B by $>$3\%. The adapted encoder representation also yields better results on SuperGLUE. We will release our checkpoints to facilitate future research.

Long-context capabilities are essential for a wide range of applications, including document and video understanding, in-context learning, and inference-time scaling, all of which require models to process and reason over long sequences of text and multimodal data. In this work, we introduce a efficient training recipe for building ultra-long context LLMs from aligned instruct model, pushing the boundaries of context lengths from 128K to 1M, 2M, and 4M tokens. Our approach leverages efficient continued pretraining strategies to extend the context window and employs effective instruction tuning to maintain the instruction-following and reasoning abilities. Our UltraLong-8B, built on Llama3.1-Instruct with our recipe, achieves state-of-the-art performance across a diverse set of long-context benchmarks. Importantly, models trained with our approach maintain competitive performance on standard benchmarks, demonstrating balanced improvements for both long and short context tasks. We further provide an in-depth analysis of key design choices, highlighting the impacts of scaling strategies and data composition. Our findings establish a robust framework for efficiently scaling context lengths while preserving general model capabilities. We release all model weights at: https://ultralong.github.io/.

Therapeutic development is a costly and high-risk endeavor that is often plagued by high failure rates. To address this, we introduce TxGemma, a suite of efficient, generalist large language models (LLMs) capable of therapeutic property prediction as well as interactive reasoning and explainability. Unlike task-specific models, TxGemma synthesizes information from diverse sources, enabling broad application across the therapeutic development pipeline. The suite includes 2B, 9B, and 27B parameter models, fine-tuned from Gemma-2 on a comprehensive dataset of small molecules, proteins, nucleic acids, diseases, and cell lines. Across 66 therapeutic development tasks, TxGemma achieved superior or comparable performance to the state-of-the-art generalist model on 64 (superior on 45), and against state-of-the-art specialist models on 50 (superior on 26). Fine-tuning TxGemma models on therapeutic downstream tasks, such as clinical trial adverse event prediction, requires less training data than fine-tuning base LLMs, making TxGemma suitable for data-limited applications. Beyond these predictive capabilities, TxGemma features conversational models that bridge the gap between general LLMs and specialized property predictors. These allow scientists to interact in natural language, provide mechanistic reasoning for predictions based on molecular structure, and engage in scientific discussions. Building on this, we further introduce Agentic-Tx, a generalist therapeutic agentic system powered by Gemini 2.5 that reasons, acts, manages diverse workflows, and acquires external domain knowledge. Agentic-Tx surpasses prior leading models on the Humanity's Last Exam benchmark (Chemistry & Biology) with 52.3% relative improvement over o3-mini (high) and 26.7% over o3-mini (high) on GPQA (Chemistry) and excels with improvements of 6.3% (ChemBench-Preference) and 2.4% (ChemBench-Mini) over o3-mini (high).

--Software requirements expressed in natural language (NL) frequently suffer from verbosity, ambiguity, and inconsistency. This creates a range of challenges, including selecting an appropriate architecture for a system and assessing different architectural alternatives. Relying on human expertise to accomplish the task of mapping NL requirements to architecture is time-consuming and error-prone. This paper proposes ARLO, an approach that automates this task by leveraging (1) a set of NL requirements for a system, (2) an existing standard that specifies architecturally relevant software quality attributes, and (3) a readily available Large Language Model (LLM). Specifically, ARLO determines the subset of NL requirements for a given system that is architecturally relevant and maps that subset to a tailorable matrix of architectural choices. ARLO applies integer linear programming on the architectural-choice matrix to determine the optimal architecture for the current requirements. We demonstrate ARLO's efficacy using a set of real-world examples. We highlight ARLO's ability (1) to trace the selected architectural choices to the requirements and (2) to isolate NL requirements that exert a particular influence on a system's architecture. This allows the identification, comparative assessment, and exploration of alternative architectural choices based on the requirements and constraints expressed therein. I NTRODUCTION Software systems are ever-evolving, with increases in size and complexity that call for careful elicitation of requirements and architectural choices [1]. While it has been long recognized that requirements and architecture co-evolve [2], [3], understanding their interactions, especially early in the software development process, is still an open challenge. More specifically, there is a scarcity of knowledge regarding their alignment, architecture-to-requirements traceability, and conserving architectural knowledge [4]. Most software requirements are still captured using natural language (NL) [5]-[7]. The informal nature of NLs is a significant obstacle in machine processing of such requirements [8]. In the past, researchers have proposed approaches to classify the NL requirements into different categories of functional and non-functional requirements [5], [6], to identify quality attributes from NL requirements, to facilitate architectural decisions by leveraging machine learning [7], and so on. While promising, these approaches require large, curated datasets and high model-building effort [9]. Translating NL requirements to architecture-related design decisions is a cumbersome task that has long been recognized in the research community [1].

LLM training is scaled up to 10Ks of GPUs by a mix of data-(DP) and model-parallel (MP) execution. Critical to achieving efficiency is tensor-parallel (TP; a form of MP) execution within tightly-coupled subsets of GPUs, referred to as a scale-up domain, and the larger the scale-up domain the better the performance. New datacenter architectures are emerging with more GPUs able to be tightly-coupled in a scale-up domain, such as moving from 8 GPUs to 72 GPUs connected via NVLink. Unfortunately, larger scale-up domains increase the blast-radius of failures, with a failure of single GPU potentially impacting TP execution on the full scale-up domain, which can degrade overall LLM training throughput dramatically. With as few as 0.1% of GPUs being in a failed state, a high TP-degree job can experience nearly 10% reduction in LLM training throughput. We propose nonuniform-tensor-parallelism (NTP) to mitigate this amplified impact of GPU failures. In NTP, a DP replica that experiences GPU failures operates at a reduced TP degree, contributing throughput equal to the percentage of still-functional GPUs. We also propose a rack-design with improved electrical and thermal capabilities in order to sustain power-boosting of scale-up domains that have experienced failures; combined with NTP, this can allow the DP replica with the reduced TP degree (i.e., with failed GPUs) to keep up with the others, thereby achieving near-zero throughput loss for large-scale LLM training.

This work presents the task of unsupervised location mapping, which seeks to map the trajectory of an individual narrative on a spatial map of locations in which a large set of narratives take place. Despite the fundamentality and generality of the task, very little work addressed the spatial mapping of narrative texts. The task consists of two parts: (1) inducing a ``map'' with the locations mentioned in a set of texts, and (2) extracting a trajectory from a single narrative and positioning it on the map. Following recent advances in increasing the context length of large language models, we propose a pipeline for this task in a completely unsupervised manner without predefining the set of labels. We test our method on two different domains: (1) Holocaust testimonies and (2) Lake District writing, namely multi-century literature on travels in the English Lake District. We perform both intrinsic and extrinsic evaluations for the task, with encouraging results, thereby setting a benchmark and evaluation practices for the task, as well as highlighting challenges.

Driving scene understanding is a critical real-world problem that involves interpreting and associating various elements of a driving environment, such as vehicles, pedestrians, and traffic signals. Despite advancements in autonomous driving, traditional pipelines rely on deterministic models that fail to capture the probabilistic nature and inherent uncertainty of real-world driving. To address this, we propose PRIMEDrive-CoT, a novel uncertainty-aware model for object interaction and Chain-of-Thought (CoT) reasoning in driving scenarios. In particular, our approach combines LiDAR-based 3D object detection with multi-view RGB references to ensure interpretable and reliable scene understanding. Uncertainty and risk assessment, along with object interactions, are modelled using Bayesian Graph Neural Networks (BGNNs) for probabilistic reasoning under ambiguous conditions. Interpretable decisions are facilitated through CoT reasoning, leveraging object dynamics and contextual cues, while Grad-CAM visualizations highlight attention regions. Extensive evaluations on the DriveCoT dataset demonstrate that PRIMEDrive-CoT outperforms state-of-the-art CoT and risk-aware models.

Large language models show promising results in various NLP tasks. Despite these successes, the robustness and consistency of LLMs in underrepresented languages remain largely unexplored, especially concerning local dialects. Existing benchmarks also focus on main dialects, neglecting LLMs' ability on local dialect texts. In this paper, we introduce a Thai local dialect benchmark covering Northern (Lanna), Northeastern (Isan), and Southern (Dambro) Thai, evaluating LLMs on five NLP tasks: summarization, question answering, translation, conversation, and food-related tasks. Furthermore, we propose a human evaluation guideline and metric for Thai local dialects to assess generation fluency and dialect-specific accuracy. Results show that LLM performance declines significantly in local Thai dialects compared to standard Thai, with only proprietary models like GPT-4o and Gemini2 demonstrating some fluency

--The Mixture of Experts (MoE) architecture has demonstrated significant advantages as it enables to increase the model capacity without a proportional increase in computation. However, the large MoE model size still introduces substantial memory demands, which usually requires expert offloading on resource-constrained platforms and incurs significant overhead. Hybrid CPU-GPU inference has been proposed to leverage CPU computation to reduce expert loading overhead but faces major challenges: on one hand, the expert activation patterns of MoE models are highly unstable, rendering the fixed mapping strategies in existing works inefficient; on the other hand, the hybrid CPU-GPU schedule for MoE is inherently complex due to the diverse expert sizes, structures, uneven workload distribution, etc. T o address these challenges, in this paper, we propose HybriMoE, a hybrid CPU-GPU inference framework that improves resource utilization through a novel CPU-GPU scheduling and cache management system. HybriMoE introduces (i) a dynamic intra-layer scheduling strategy to balance workloads across CPU and GPU, (ii) an impact-driven inter-layer prefetching algorithm, and (iii) a score-based caching algorithm to mitigate expert activation instability. We implement HybriMoE on top of the kTransformers framework and evaluate it on three widely used MoE-based LLMs. Experimental results demonstrate that HybriMoE achieves an average speedup of 1.33 in the prefill stage and 1.70 in the decode stage compared to state-of-the-art hybrid MoE inference framework. Mixture of Experts (MoE) has emerged as a promising solution to enhance computational efficiency of Large Language Models (LLMs) without compromising model performance [1], [2].

Large language models have made significant advancements in various natural language processing tasks, including coreference resolution. However, traditional methods often fall short in effectively distinguishing referential relationships due to a lack of integration between syntactic and semantic information. This study introduces an innovative framework aimed at enhancing coreference resolution by utilizing pretrained language models. Our approach combines syntax parsing with semantic role labeling to accurately capture finer distinctions in referential relationships. By employing state-of-the-art pretrained models to gather contextual embeddings and applying an attention mechanism for fine-tuning, we improve the performance of coreference tasks. Experimental results across diverse datasets show that our method surpasses conventional coreference resolution systems, achieving notable accuracy in disambiguating references. This development not only improves coreference resolution outcomes but also positively impacts other natural language processing tasks that depend on precise referential understanding.