We study reasoning tasks through a framework that integrates auto-regressive (AR) and non-autoregressive (NAR) language models. AR models, which generate text sequentially, excel at producing coherent outputs but often suffer from slow inference, particularly in reasoning-intensive domains such as mathematics and code, where lengthy chains of thought are required. In contrast, NAR models, such as discrete diffusion models, allow parallel generation and offer substantial speedups, though typically at the cost of reduced output quality. To address these limitations, we introduce a new paradigm in which an NAR model efficiently produces intermediate reasoning traces, which subsequently guide an AR model to deliver precise final answers. Experiments demonstrate that our approach yields significant 26% improvements over strong baselines while substantially reducing inference cost.

Variational quantum algorithms (VQAs) are leading strategies to reach practical utilities of near-term quantum devices. However, the no-cloning theorem in quantum mechanics precludes standard backpropagation, leading to prohibitive quantum resource costs when applying VQAs to large-scale tasks. To address this challenge, we reformulate the training dynamics of VQAs as a nonlinear partial differential equation and propose a novel protocol that leverages physics-informed neural networks (PINNs) to model this dynamical system efficiently. Given a small amount of training trajectory data collected from quantum devices, our protocol predicts the parameter updates of VQAs over multiple iterations on the classical side, dramatically reducing quantum resource costs. Through systematic numerical experiments, we demonstrate that our method achieves up to a 30x speedup compared to conventional methods and reduces quantum resource costs by as much as 90\% for tasks involving up to 40 qubits, including ground state preparation of different quantum systems, while maintaining competitive accuracy. Our approach complements existing techniques aimed at improving the efficiency of VQAs and further strengthens their potential for practical applications.

As designers become familiar with Generative AI, a new concept is emerging: Agentic AI. While generative AI produces output in response to prompts, agentic AI systems promise to perform mundane tasks autonomously, potentially freeing designers to focus on what they love: being creative. But how do designers feel about integrating agentic AI systems into their workflows? Through design fiction, we investigated how designers want to interact with a collaborative agentic AI platform. Ten professional designers imagined and discussed collaborating with an AI agent to organise inspiration sources and ideate. Our findings highlight the roles AI agents can play in supporting designers, the division of authority between humans and AI, and how designers' intent can be explained to AI agents beyond prompts. We synthesise our findings into a conceptual framework that identifies authority distribution among humans and AI agents and discuss directions for utilising AI agents in future design workflows.

Inspired by the effectiveness of genetic algorithms and the importance of synthesizability in molecular design, we present SynGA, a simple genetic algorithm that operates directly over synthesis routes. Our method features custom crossover and mutation operators that explicitly constrain it to synthesizable molecular space. By modifying the fitness function, we demonstrate the effectiveness of SynGA on a variety of design tasks, including synthesizable analog search and sample-efficient property optimization, for both 2D and 3D objectives. Furthermore, by coupling SynGA with a machine learning-based filter that focuses the building block set, we boost SynGA to state-of-the-art performance. For property optimization, this manifests as a model-based variant SynGBO, which employs SynGA and block filtering in the inner loop of Bayesian optimization. Since SynGA is lightweight and enforces synthesizability by construction, our hope is that SynGA can not only serve as a strong standalone baseline but also as a versatile module that can be incorporated into larger synthesis-aware workflows in the future.

ABSTRACT The adoption of cyber-physical systems and jobsite intelligence that connects design models, real-time site sensing, and autonomous field operations can dramatically enhance digital management in the Architecture, Engineering, and Construction (AEC) industry. This paper introduces BIM2RDT (Building Information Models to Robot-Ready Site Digital T wins), an agentic artificial intelligence (AI) framework designed to transform static Building Information Modeling (BIM) into dynamic, robot-ready digital twins (DTs) that prioritize safety during construction execution. The framework bridges the gap between pre-existing BIM data and real-time site conditions by integrating three key data streams: geometric and semantic information from BIM models, real-time activity data from IoT sensor networks, and visual-spatial data collected by quadruped robots during site traversal. The methodology introduces Semantic-Gravity ICP (SG-ICP), a novel point cloud registration algorithm that leverages large language model (LLM) reasoning. This creates an intelligent feedback loop where robot-collected data updates the DT, which in turn optimizes paths for subsequent missions. The framework employs YOLOE open-vocabulary object detection and Shi-Tomasi corner detection to identify and track construction elements while using BIM geometry as robust a priori maps. Major findings from experiments demonstrate SG-ICP's superiority over standard ICP, achieving RMSE reductions of 64.3%-88.3% in alignment across varied scenarios with occluded or sparse features, ensuring physically plausible orientations. HA V integration triggers real-time warnings and tasks upon exceeding exposure limits, enhancing compliance with such standards as ISO 5349-1. PRACTICAL APPLICATIONS Construction sites are becoming increasingly complex with the introduction of new technologies such as reality capture equipment and robots, requiring better tools to streamline adoption, avoid tool sprawl, and ensure worker safety. This research introduces a system that combines robots, smart sensors, and building information modeling (BIM) data to create a "digital twin": an up-to-date virtual copy of a construction site's geometries and safety information. The system uses quadruped robots equipped with cameras and sensors to autonomously walk through construction sites, automatically detecting and tracking objects like equipment, materials, and temporary structures. Unlike traditional approaches that start from scratch, this method leverages existing BIM data as a foundation, making the robots more accurate and efficient at understanding their surroundings. Besides geometric site updates, safety information is also presented in the updated digital twin.

In the past few years, foundation models based on large transformer networks such as Google's BERT (Kenton and Toutanova, 2019) and OpenAI's GPT family (Radford, 2018) have been shown to be invaluable aids for scientific discovery in the analysis of genomic data (Cui et al., 2024; Theodoris et al., 2023; Chen and Zou, 2025). More specifically, foundation models targeted for genomic applications typically comprise of those that are trained on enormous databases of experimental data such as scGPT (Cui et al., 2024), which was trained on transcriptomes from 33 million human cells from 441 different studies or the GeneFormer model (Theodoris et al., 2023), which was trained on 29.9 million human single-cell transcriptomes. On the other hand, foundation models based on pre-training on internet-scale databases of natural language texts may offer distinct advantages, such as potentially taking advantage of niche biological relationships which may be widely documented in scientific literature, but not necessarily be represented experimentally in large-scale genomics datasets. For this reason, some recent works have used the embedding outputs of large-language models (LLMs) such as ChatGPT (Radford, 2018) to encode the biological information contained in text-based gene descriptions, such as those in the NCBI database (Schoch et al., 2020). Notably, Chen and Zou (2025) show that these text-based gene descriptors can be input to GPT-3.5 to obtain gene embeddings that act as features/covariates for standard prediction algorithms, denoted GenePT.

Accurate prediction of drug-target binding affinity can accelerate drug discovery by prioritizing promising compounds before costly wet-lab screening. While deep learning has advanced this task, most models fuse ligand and protein representations via simple concatenation and lack explicit geometric regularization, resulting in poor generalization across chemical space and time. We introduce FIRM-DTI, a lightweight framework that conditions molecular embeddings on protein embeddings through a feature-wise linear modulation (FiLM) layer and enforces metric structure with a triplet loss. An RBF regression head operating on embedding distances yields smooth, interpretable affinity predictions. Despite its modest size, FIRM-DTI achieves state-of-the-art performance on the Therapeutics Data Commons DTI-DG benchmark, as demonstrated by an extensive ablation study and out-of-domain evaluation. Our results underscore the value of conditioning and metric learning for robust drug-target affinity prediction.

Neural architecture search (NAS) has shown great promise in automatically designing lightweight models. However, conventional approaches are insufficient in training the supernet and pay little attention to actual robot hardware resources. To meet such challenges, we propose RAM-NAS, a resource-aware multi-objective NAS method that focuses on improving the supernet pretrain and resource-awareness on robot hardware devices. We introduce the concept of subnets mutual distillation, which refers to mutually distilling all subnets sampled by the sandwich rule. Additionally, we utilize the Decoupled Knowledge Distillation (DKD) loss to enhance logits distillation performance. To expedite the search process with consideration for hardware resources, we used data from three types of robotic edge hardware to train Latency Surrogate predictors. These predictors facilitated the estimation of hardware inference latency during the search phase, enabling a unified multi-objective evolutionary search to balance model accuracy and latency trade-offs. Our discovered model family, RAM-NAS models, can achieve top-1 accuracy ranging from 76.7% to 81.4% on ImageNet. In addition, the resource-aware multi-objective NAS we employ significantly reduces the model's inference latency on edge hardware for robots. We conducted experiments on downstream tasks to verify the scalability of our methods. The inference time for detection and segmentation is reduced on all three hardware types compared to MobileNetv3-based methods. Our work fills the gap in NAS for robot hardware resource-aware.

Fine-tuning large language models (LLMs) with local data is a widely adopted approach for organizations seeking to adapt LLMs to their specific domains. Given the shared characteristics in data across different organizations, the idea of collaboratively fine-tuning an LLM using data from multiple sources presents an appealing opportunity. However, organizations are often reluctant to share local data, making centralized fine-tuning impractical. Federated learning (FL), a privacy-preserving framework, enables clients to retain local data while sharing only model parameters for collaborative training, offering a potential solution. While fine-tuning LLMs on centralized datasets risks data leakage through next-token prediction, the iterative aggregation process in FL results in a global model that encapsulates generalized knowledge, which some believe protects client privacy. In this paper, however, we present contradictory findings through extensive experiments. We show that attackers can still extract training data from the global model, even using straightforward generation methods, with leakage increasing as the model size grows. Moreover, we introduce an enhanced attack strategy tailored to FL, which tracks global model updates during training to intensify privacy leakage. To mitigate these risks, we evaluate privacy-preserving techniques in FL, including differential privacy, regularization-constrained updates and adopting LLMs with safety alignment. Our results provide valuable insights and practical guidelines for reducing privacy risks when training LLMs with FL.

In this work, we develop machine learning (ML) based strategies to predict resources (costs) required for massively parallel chemistry computations, such as coupled-cluster methods, to guide application users before they commit to running expensive experiments on a supercomputer. By predicting application execution time, we determine the optimal runtime parameter values such as number of nodes and tile sizes. Two key questions of interest to users are addressed. The first is the shortest-time question, where the user is interested in knowing the parameter configurations (number of nodes and tile sizes) to achieve the shortest execution time for a given problem size and a target supercomputer. The second is the cheapest-run question in which the user is interested in minimizing resource usage, i.e., finding the number of nodes and tile size that minimizes the number of node-hours for a given problem size. We evaluate a rich family of ML models and strategies, developed based on the collections of runtime parameter values for the CCSD (Coupled Cluster with Singles and Doubles) application executed on the Department of Energy (DOE) Frontier and Aurora supercomputers. Our experiments show that when predicting the total execution time of a CCSD iteration, a Gradient Boosting (GB) ML model achieves a Mean Absolute Percentage Error (MAPE) of 0.023 and 0.073 for Aurora and Frontier, respectively. In the case where it is expensive to run experiments just to collect data points, we show that active learning can achieve a MAPE of about 0.2 with just around 450 experiments collected from Aurora and Frontier.