Prior benchmarks for evaluating the domain-specific knowledge of large language models (LLMs) lack the scalability to handle complex academic tasks. To address this, we introduce \texttt{ScholarBench}, a benchmark centered on deep expert knowledge and complex academic problem-solving, which evaluates the academic reasoning ability of LLMs and is constructed through a three-step process. \texttt{ScholarBench} targets more specialized and logically complex contexts derived from academic literature, encompassing five distinct problem types. Unlike prior benchmarks, \texttt{ScholarBench} evaluates the abstraction, comprehension, and reasoning capabilities of LLMs across eight distinct research domains. To ensure high-quality evaluation data, we define category-specific example attributes and design questions that are aligned with the characteristic research methodologies and discourse structures of each domain. Additionally, this benchmark operates as an English-Korean bilingual dataset, facilitating simultaneous evaluation for linguistic capabilities of LLMs in both languages. The benchmark comprises 5,031 examples in Korean and 5,309 in English, with even state-of-the-art models like o3-mini achieving an average evaluation score of only 0.543, demonstrating the challenging nature of this benchmark.

Plans for a groundbreaking rare earths refinery in East Yorkshire have been abandoned, after the company behind the project decided to seek investment in the United States instead. Pensana has spent the past seven years developing a rare earths mine in Angola. The $268m (£185m) project, one of the largest of its kind in the world, will begin delivering raw materials in 2027. The company had planned to build a refinery at the Saltend Chemicals Plant near Hull, which would have processed the raw materials into metals used to create powerful magnets. These magnets would then be used in high-tech applications such as motors for electric vehicles, wind turbines and robotics.

Although digital fabrication processes at the desktop scale have become proficient and prolific, systems aimed at producing larger-scale structures are still typically complex, expensive, and unreliable. In this work, we present an approach for the fabrication of scalable macroscale structures using simple robots and interlocking lattice building blocks. A target structure is first voxelized so that it can be populated with an architected lattice. These voxels are then grouped into larger interconnected blocks, which are produced using standard digital fabrication processes, leveraging their capability to produce highly complex geometries at a small scale. These blocks, on the size scale of tens of centimeters, are then fed to mobile relative robots that are able to traverse over the structure and place new blocks to form structures on the meter scale. To facilitate the assembly of large structures, we introduce a live digital twin simulation tool for controlling and coordinating assembly robots that enables both global planning for a target structure and live user design, interaction, or intervention. To improve assembly throughput, we introduce a new modular assembly robot, designed for hierarchical voxel handling. We validate this system by demonstrating the voxelization, hierarchical blocking, path planning, and robotic fabrication of a set of meter-scale objects.

LiDAR Simultaneous Localization and Mapping (SLAM) systems are essential for enabling precise navigation and environmental reconstruction across various applications. Although current point-to-plane ICP algorithms perform effec- tively in structured, feature-rich environments, they struggle in scenarios with sparse features, repetitive geometric structures, and high-frequency motion. This leads to degeneracy in 6- DOF pose estimation. Most state-of-the-art algorithms address these challenges by incorporating additional sensing modalities, but LiDAR-only solutions continue to face limitations under such conditions. To address these issues, we propose a novel Degeneracy-Aware Multi-Metric LiDAR Odometry and Map- ping (DAMM-LOAM) module. Our system improves mapping accuracy through point cloud classification based on surface normals and neighborhood analysis. Points are classified into ground, walls, roof, edges, and non-planar points, enabling accurate correspondences. A Degeneracy-based weighted least squares-based ICP algorithm is then applied for accurate odom- etry estimation. Additionally, a Scan Context based back-end is implemented to support robust loop closures. DAMM-LOAM demonstrates significant improvements in odometry accuracy, especially in indoor environments such as long corridors

Real-world financial documents report essential information about an entity's financial holdings that can span millions of different financial instrument types. Yet, these details are often buried in messy, multi-page, fragmented tables - for example, 99.4% of the tables in our dataset have no bounding boxes with the maximum number of rows amounting to 426 per table across 44 pages. To tackle these unique challenges from real-world tables, we present a continuously learning, agentic table extraction system, TASER (Table Agents for Schema-guided Extraction and Recommendation) that extracts highly unstructured, multi-page, heterogeneous tables into normalized, schema-conforming outputs. Our table agents execute on table detection, classification, extraction, and recommendations by leveraging an initial schema. Then, our Recommender Agent reviews the outputs, recommends schema revisions, and decides on the final recommendations, enabling TASER to outperform existing table detection models such as Table Transformer by 10.1%. Within this continuous learning process, we highlight that larger batch sizes result in a 104.3% increase in schema recommendations that are actionable and utilized, resulting in a 9.8% increase in extracted holdings - highlighting the importance of a continuous learning process. To train TASER, we have manually labeled 22,584 pages (28,150,449 tokens), 3,213 tables for $731,685,511,687 of holdings culminating in one of the first real financial table datasets. We release our dataset TASERTab to enable the research community to access real-world financial tables and outputs. Our results highlight the promise of agentic, schema-guided extraction systems for robust understanding of real-world financial tables.

Reaction mechanism search tools have demonstrated the ability to provide insights into likely products and rate-limiting steps of reacting systems. However, reactions involving several concerted bond changes - as can be found in many key steps of natural product synthesis - can complicate the search process. To mitigate these complications, we present a mechanism search strategy particularly suited to help expedite exploration of an exemplary family of such complex reactions, cyclizations. We provide a cost-effective strategy for identifying relevant elementary reaction steps by combining graph-based enumeration schemes and machine learning techniques for intermediate filtering. Key to this approach is our use of a neural network potential (NNP), AIMNet2-rxn, for computational evaluation of each candidate reaction pathway. In this article, we evaluate the NNP's ability to estimate activation energies, demonstrate the correct anticipation of stereoselectivity, and recapitulate complex enabling steps in natural product synthesis.

Effectively integrating molecular graph structures with Large Language Models (LLMs) is a key challenge in drug discovery. Most existing multi-modal alignment methods typically process these structures by fine-tuning the LLM or adding a static adapter simultaneously. However, these approaches have two main limitations: (1) it optimizes a shared parameter space across all molecular inputs, limiting the model's ability to capture instance-specific structural features; and (2) fine-tuning the LLM for molecular tasks can lead to catastrophic forgetting, undermining its general reasoning capabilities. In this paper, instead of static task-oriented adaptation, we propose an instance-specific parameter space alignment approach for each molecule on-the-fly. To this end, we introduce Molecule-aware Low-Rank Adaptation (MoRA) that produces a unique set of low-rank adaptation weights for each input molecular graph. These weights are then dynamically injected into a frozen LLM, allowing the model to adapt its reasoning to the structure of each molecular input, while preserving the LLM's core knowledge. Extensive experiments demonstrate that on key molecular tasks, such as chemical reaction prediction and molecular captioning, MoRA's instance-specific dynamic adaptation outperforms statically adapted baselines, including a 14.1% relative improvement in reaction prediction exact match and a 22% reduction in error for quantum property prediction. The code is available at https://github.com/jk-sounds/MoRA.

Multilingual domain adaptation (ML-DA) is widely used to learn new domain knowledge across languages into large language models (LLMs). Although many methods have been proposed to improve domain adaptation, the mechanisms of multilingual knowledge acquisition, how domain knowledge is learned within a language and transferred across languages, remain underexplored. This gap leads to suboptimal performance, particularly in low-resource settings. This work examines the learning dynamics of LLMs during ML-DA. Because prior ML-DA studies often train and evaluate on datasets with mismatched knowledge coverage, we propose AdaXEval, an adaptive evaluation method that builds multiple-choice QA datasets from the same bilingual domain corpus used for training, thereby directly studying multilingual knowledge acquisition. Through continual training of LLMs with diverse data recipes, we track how LLMs acquire domain facts and pinpoint the mechanism behind the transformation process from domain training data to knowledge. Our experiments on a 13B English-Japanese bilingual LLM reveal that cross-lingual transfer remains challenging despite a high-quality bilingual corpus. The code has been released.

It looks like fabric, feels like metal, and is as light as rubber. Any watch fan looking to tick all of the above boxes would normally expect to be a dab hand with a spring bar removal tool to experience all the above individually, but a new strap developed by Malaysian independent brand Ming appears to now offer the best of all worlds. The one strap to rule them all has been dubbed the Polymesh, and is 3D-printed from grade five titanium, and comprises 1,693 interconnected pieces (including the buckle) held together without any pins or screws. The only additional parts requiring assembly are the quick-release spring bars at each end that attach it to the watch--the articulated pin buckle is also formed in the same process. Ming says that the strap, which is made up from rows of 15 equilateral triangles, meshed together and bookended by larger end pieces, "has more motion engineered into the radial axis than the lateral one," leading to a supple end result that drapes like fabric yet retains the strength of titanium.

We introduce $\textbf{MASSE}$, the first Multi-Agent System for Structural Engineering, effectively integrating large language model (LLM)-based agents with real-world engineering workflows. Structural engineering is a fundamental yet traditionally stagnant domain, with core workflows remaining largely unchanged for decades despite its substantial economic impact and global market size. Recent advancements in LLMs have significantly enhanced their ability to perform complex reasoning, long-horizon planning, and precise tool utilization -- capabilities well aligned with structural engineering tasks such as interpreting design codes, executing load calculations, and verifying structural capacities. We present a proof-of-concept showing that most real-world structural engineering workflows can be fully automated through a training-free LLM-based multi-agent system. MASSE enables immediate deployment in professional environments, and our comprehensive validation on real-world case studies demonstrates that it can reduce expert workload from approximately two hours to mere minutes, while enhancing both reliability and accuracy in practical engineering scenarios.