Fine-tuning large language models (LLMs) often faces GPU memory bottlenecks: the backward pass of first-order optimizers like Adam increases memory usage to more than 10 times the inference level (e.g., 633 GB for OPT-30B). Zeroth-order (ZO) optimizers avoid this cost by estimating gradients only from forward passes, yet existing methods like MeZO usually require many more steps to converge. Can this trade-off between speed and memory in ZO be fundamentally improved? Normalized-SGD demonstrates strong empirical performance with greater memory efficiency than Adam. In light of this, we introduce FZOO, a Fast Zeroth-Order Optimizer toward Adam-Scale Speed. FZOO reduces the total forward passes needed for convergence by employing batched one-sided estimates that adapt step sizes based on the standard deviation of batch losses. It also accelerates per-batch computation through the use of Rademacher random vector perturbations coupled with CUDA's parallel processing. Extensive experiments on diverse models, including RoBERTa-large, OPT (350M-66B), Phi-2, and Llama3, across 11 tasks validate FZOO's effectiveness. On average, FZOO outperforms MeZO by 3 percent in accuracy while requiring 3 times fewer forward passes. For RoBERTa-large, FZOO achieves average improvements of 5.6 percent in accuracy and an 18 times reduction in forward passes compared to MeZO, achieving convergence speeds comparable to Adam. We also provide theoretical analysis proving FZOO's formal equivalence to a normalized-SGD update rule and its convergence guarantees. FZOO integrates smoothly into PEFT techniques, enabling even larger memory savings. Overall, our results make single-GPU, high-speed, full-parameter fine-tuning practical and point toward future work on memory-efficient pre-training.

A significant portion of the energy consumed by Large Language Models (LLMs) arises from their inference processes; hence developing energy-efficient methods for inference is crucial. While several techniques exist for inference optimization, output compression remains relatively unexplored, with only a few preliminary efforts addressing this aspect. In this work, we first benchmark 12 decoder-only LLMs across 5 datasets, revealing that these models often produce responses that are substantially longer than necessary. We then conduct a comprehensive quality assessment of LLM responses, formally defining six information categories present in LLM responses. We show that LLMs often tend to include redundant or additional information besides the minimal answer. To address this issue of long responses by LLMs, we explore several simple and intuitive prompt-engineering strategies. Empirical evaluation shows that appropriate prompts targeting length reduction and controlling information content can achieve significant energy optimization between 25-60\% by reducing the response length while preserving the quality of LLM responses.

Modern chip design is complex, and there is a crucial need for early-stage prediction of key design-quality metrics like timing and routing congestion directly from Verilog code (a commonly used programming language for hardware design). It is especially important yet complex to predict individual lines of code that cause timing violations or downstream routing congestion. Prior works have tried approaches like converting Verilog into an intermediate graph representation and using LLM embeddings alongside other features to predict module-level quality, but did not consider line-level quality prediction. We propose VeriLoC, the first method that predicts design quality directly from Verilog at both the line- and module-level. To this end, VeriLoC leverages recent Verilog code-generation LLMs to extract local line-level and module-level embeddings, and train downstream classifiers/regressors on concatenations of these embeddings. VeriLoC achieves high F1-scores of 0.86-0.95 for line-level congestion and timing prediction, and reduces the mean average percentage error from 14% - 18% for SOTA methods down to only 4%. We believe that VeriLoC embeddings and insights from our work will also be of value for other predictive and optimization tasks for complex hardware design.

Existing computer-use agents primarily focus on general-purpose desktop automation tasks, with limited exploration of their application in highly specialized domains. In particular, the 3D building modeling process in the Architecture, Engineering, and Construction (AEC) sector involves open-ended design tasks and complex interaction patterns within Building Information Modeling (BIM) authoring software, which has yet to be thoroughly addressed by current studies. In this paper, we propose BIMgent, an agentic framework powered by multimodal large language models (LLMs), designed to enable autonomous building model authoring via graphical user interface (GUI) operations. BIMgent automates the architectural building modeling process, including multimodal input for conceptual design, planning of software-specific workflows, and efficient execution of the authoring GUI actions. We evaluate BIMgent on real-world building modeling tasks, including both text-based conceptual design generation and reconstruction from existing building design. The design quality achieved by BIMgent was found to be reasonable. Its operations achieved a 32% success rate, whereas all baseline models failed to complete the tasks (0% success rate). Results demonstrate that BIMgent effectively reduces manual workload while preserving design intent, highlighting its potential for practical deployment in real-world architectural modeling scenarios. Project page: https://tumcms.github.io/BIMgent.github.io/

Autonomous UAV operation necessitates reliable mathematical reasoning for tasks such as trajectory planning and power management. While traditional flight control relies on hardcoded equations, recent Large Language Models (LLMs) offer potential for more flexible problem-solving but struggle with reliably selecting and applying correct mathematical formulations and executing precise multi-step arithmetic. We propose RAG-UAV, a retrieval-augmented generation framework designed to improve the mathematical reasoning of several LLMs (including GPT o1/Turbo, Llama-3.2/3.3, Mistral, and DeepSeek R1) in UAV-specific contexts by providing access to relevant domain literature. To conduct an initial assessment, we introduce the UAV-Math-Bench, a 20-question problem set of UAV-centric mathematical problems across four difficulty levels. Our experiments demonstrate that incorporating retrieval substantially increases exact answer accuracy (achieving up to 75% with o1), reduces instances of incorrect formulation selection (from 25% without RAG to 5\% with RAG), and decreases numerical errors, reducing Mean Squared Error (MSE) by orders of magnitude for the best-performing models. This pilot study indicates that RAG can enable general-purpose LLMs to function as more reliable tools for engineering analysis, although direct real-time flight control requires further investigation and validation on a larger scale. All benchmark data, questions, and answers are publicly available.

This paper introduces a dual-layer framework for network operator-side quality of experience (QoE) assessment that integrates both objective network modeling and subjective user perception extracted from live-streaming platforms. On the objective side, we develop a machine learning model trained on mean opinion scores (MOS) computed via the ITU-T P.1203 reference implementation, allowing accurate prediction of user-perceived video quality using only network parameters such as packet loss, delay, jitter, and throughput without reliance on video content or client-side instrumentation. On the subjective side, we present a semantic filtering and scoring pipeline that processes user comments from live streams to extract performance-related feedback. A large language model is used to assign scalar MOS scores to filtered comments in a deterministic and reproducible manner. To support scalable and interpretable analysis, we construct a labeled dataset of 47,894 live-stream comments, of which about 34,000 are identified as QoE-relevant through multi-layer semantic filtering. Each comment is enriched with simulated Internet Service Provider attribution and temporally aligned using synthetic timestamps in 5-min intervals. The resulting dataset enables operator-level aggregation and time-series analysis of user-perceived quality. A delta MOS metric is proposed to measure each Internet service provider's deviation from platform-wide sentiment, allowing detection of localized degradations even in the absence of direct network telemetry. A controlled outage simulation confirms the framework's effectiveness in identifying service disruptions through comment-based trends alone. The system provides each operator with its own subjective MOS and the global platform average per interval, enabling real-time interpretation of performance deviations and comparison with objective network-based QoE estimates.

Large Language Models (LLMs) are increasingly used to support scientific research, but their knowledge of scientific advancements can quickly become outdated. We introduce ScienceMeter, a new framework for evaluating scientific knowledge update methods over scientific knowledge spanning the past, present, and future. ScienceMeter defines three metrics: knowledge preservation, the extent to which models' understanding of previously learned papers are preserved; knowledge acquisition, how well scientific claims from newly introduced papers are acquired; and knowledge projection, the ability of the updated model to anticipate or generalize to related scientific claims that may emerge in the future. Using ScienceMeter, we examine the scientific knowledge of LLMs on claim judgment and generation tasks across a curated dataset of 15,444 scientific papers and 30,888 scientific claims from ten domains including medicine, biology, materials science, and computer science. We evaluate five representative knowledge update approaches including training- and inference-time methods. With extensive experiments, we find that the best-performing knowledge update methods can preserve only 85.9% of existing knowledge, acquire 71.7% of new knowledge, and project 37.7% of future knowledge. Inference-based methods work for larger models, whereas smaller models require training to achieve comparable performance. Cross-domain analysis reveals that performance on these objectives is correlated. Even when applying on specialized scientific LLMs, existing knowledge update methods fail to achieve these objectives collectively, underscoring that developing robust scientific knowledge update mechanisms is both crucial and challenging.

This white paper discusses and explores the various points of intersection between quantum computing and artificial intelligence (AI). It describes how quantum computing could support the development of innovative AI solutions. It also examines use cases of classical AI that can empower research and development in quantum technologies, with a focus on quantum computing and quantum sensing. The purpose of this white paper is to provide a long-term research agenda aimed at addressing foundational questions about how AI and quantum computing interact and benefit one another. It concludes with a set of recommendations and challenges, including how to orchestrate the proposed theoretical work, align quantum AI developments with quantum hardware roadmaps, estimate both classical and quantum resources - especially with the goal of mitigating and optimizing energy consumption - advance this emerging hybrid software engineering discipline, and enhance European industrial competitiveness while considering societal implications.

Fine-tuning pre-trained generative models with Reinforcement Learning (RL) has emerged as an effective approach for aligning outputs more closely with nuanced human preferences. In this paper, we investigate the application of Group Relative Policy Optimization (GRPO) to fine-tune next-scale visual autoregressive (VAR) models. Our empirical results demonstrate that this approach enables alignment to intricate reward signals derived from aesthetic predictors and CLIP embeddings, significantly enhancing image quality and enabling precise control over the generation style. Interestingly, by leveraging CLIP, our method can help VAR models generalize beyond their initial ImageNet distribution: through RL-driven exploration, these models can generate images aligned with prompts referencing image styles that were absent during pre-training. In summary, we show that RL-based fine-tuning is both efficient and effective for VAR models, benefiting particularly from their fast inference speeds, which are advantageous for online sampling, an aspect that poses significant challenges for diffusion-based alternatives.

Large language models (LLMs) have demonstrated strong capabilities in translating natural language questions about relational databases into SQL queries. In particular, test-time scaling techniques such as Self-Consistency and Self-Correction can enhance SQL generation accuracy by increasing computational effort during inference. However, these methods have notable limitations: Self-Consistency may select suboptimal outputs despite majority votes, while Self-Correction typically addresses only syntactic errors. To leverage the strengths of both approaches, we propose CSC-SQL, a novel method that integrates Self-Consistency and Self-Correction. CSC-SQL selects the two most frequently occurring outputs from parallel sampling and feeds them into a merge revision model for correction. Additionally, we employ the Group Relative Policy Optimization (GRPO) algorithm to fine-tune both the SQL generation and revision models via reinforcement learning, significantly enhancing output quality. Experimental results confirm the effectiveness and generalizability of CSC-SQL. On the BIRD private test set, our 7B model achieves 71.72\% execution accuracy, while the 32B model achieves 73.67\%. The code has been open sourced at https://github.com/CycloneBoy/csc_sql.