Reasoning large language models (LLMs) have demonstrated superior capacities in solving complicated problems by generating long chain-of-thoughts (CoT), but such a lengthy CoT incurs high inference costs. In this study, we introduce ES-CoT, an inference-time method that shortens CoT generation by detecting answer convergence and stopping early with minimal performance loss. At the end of each reasoning step, we prompt the LLM to output its current final answer, denoted as a step answer. We then track the run length of consecutive identical step answers as a measure of answer convergence. Once the run length exhibits a sharp increase and exceeds a minimum threshold, the generation is terminated. We provide both empirical and theoretical support for this heuristic: step answers steadily converge to the final answer, and large run-length jumps reliably mark this convergence. Experiments on five reasoning datasets across three LLMs show that ES-CoT reduces the number of inference tokens by about 41\% on average while maintaining accuracy comparable to standard CoT. Further, ES-CoT integrates seamlessly with self-consistency prompting and remains robust across hyperparameter choices, highlighting it as a practical and effective approach for efficient reasoning.

The Universality Hypothesis in large language models (LLMs) claims that different models converge towards similar concept representations in their latent spaces. Providing evidence for this hypothesis would enable researchers to exploit universal properties, facilitating the generalization of mechanistic interpretability techniques across models. Previous works studied if LLMs learned the same features, which are internal representations that activate on specific concepts. Since comparing features across LLMs is challenging due to polysemanticity, in which LLM neurons often correspond to multiple unrelated features rather than to distinct concepts, sparse autoencoders (SAEs) have been employed to disentangle LLM neurons into SAE features corresponding to distinct concepts. In this paper, we introduce a new variation of the universality hypothesis called Analogous Feature Universality: we hypothesize that even if SAEs across different models learn different feature representations, the spaces spanned by SAE features are similar, such that one SAE space is similar to another SAE space under rotation-invariant transformations. Evidence for this hypothesis would imply that interpretability techniques related to latent spaces, such as steering vectors, may be transferred across models via certain transformations. To investigate this hypothesis, we first pair SAE features across different models via activation correlation, and then measure spatial relation similarities between paired features via representational similarity measures, which transform spaces into representations that reveal hidden relational similarities. Our experiments demonstrate high similarities for SAE feature spaces across various LLMs, providing evidence for feature space universality.

Scientific idea generation has been extensively studied in creativity theory and computational creativity research, providing valuable frameworks for understanding and implementing creative processes. However, recent work using Large Language Models (LLMs) for research idea generation often overlooks these theoretical foundations. We present a framework that explicitly implements combinatorial creativity theory using LLMs, featuring a generalization-level retrieval system for cross-domain knowledge discovery and a structured combinatorial process for idea generation. The retrieval system maps concepts across different abstraction levels to enable meaningful connections between disparate domains, while the combinatorial process systematically analyzes and recombines components to generate novel solutions. Experiments on the OAG-Bench dataset demonstrate our framework's effectiveness, consistently outperforming baseline approaches in generating ideas that align with real research developments (improving similarity scores by 7\%-10\% across multiple metrics). Our results provide strong evidence that LLMs can effectively realize combinatorial creativity when guided by appropriate theoretical frameworks, contributing both to practical advancement of AI-assisted research and theoretical understanding of machine creativity.

Scientific discoveries are often made by finding a pattern or object that was not predicted by the known rules of science. Oftentimes, these anomalous events or objects that do not conform to the norms are an indication that the rules of science governing the data are incomplete, and something new needs to be present to explain these unexpected outliers. The challenge of finding anomalies can be confounding since it requires codifying a complete knowledge of the known scientific behaviors and then projecting these known behaviors on the data to look for deviations. When utilizing machine learning, this presents a particular challenge since we require that the model not only understands scientific data perfectly but also recognizes when the data is inconsistent and out of the scope of its trained behavior. In this paper, we present three datasets aimed at developing machine learning-based anomaly detection for disparate scientific domains covering astrophysics, genomics, and polar science. We present the different datasets along with a scheme to make machine learning challenges around the three datasets findable, accessible, interoperable, and reusable (FAIR). Furthermore, we present an approach that generalizes to future machine learning challenges, enabling the possibility of large, more compute-intensive challenges that can ultimately lead to scientific discovery.

Autism spectrum disorder (ASD) is a developmental disorder characterized by significant social communication impairments and difficulties perceiving and presenting communication cues. Machine learning techniques have been broadly adopted to facilitate autism studies and assessments. However, computational models are primarily concentrated on specific analysis and validated on private datasets in the autism community, which limits comparisons across models due to privacy-preserving data sharing complications. This work presents a novel privacy-preserving open-source dataset, MMASD as a MultiModal ASD benchmark dataset, collected from play therapy interventions of children with Autism. MMASD includes data from 32 children with ASD, and 1,315 data samples segmented from over 100 hours of intervention recordings. To promote public access, each data sample consists of four privacy-preserving modalities of data; some of which are derived from original videos: (1) optical flow, (2) 2D skeleton, (3) 3D skeleton, and (4) clinician ASD evaluation scores of children, e.g., ADOS scores. MMASD aims to assist researchers and therapists in understanding children's cognitive status, monitoring their progress during therapy, and customizing the treatment plan accordingly. It also has inspiration for downstream tasks such as action quality assessment and interpersonal synchrony estimation. MMASD dataset can be easily accessed at this https URL.

Autism spectrum disorder (ASD) is characterized by significant challenges in social interaction and comprehending communication signals. Recently, therapeutic interventions for ASD have increasingly utilized Deep learning powered-computer vision techniques to monitor individual progress over time. These models are trained on private, non-public datasets from the autism community, creating challenges in comparing results across different models due to privacy-preserving data-sharing issues. This work introduces MMASD+, an enhanced version of the novel open-source dataset called Multimodal ASD (MMASD). MMASD+ consists of diverse data modalities, including 3D-Skeleton, 3D Body Mesh, and Optical Flow data. It integrates the capabilities of Yolov8 and Deep SORT algorithms to distinguish between the therapist and children, addressing a significant barrier in the original dataset. Additionally, a Multimodal Transformer framework is proposed to predict 11 action types and the presence of ASD. This framework achieves an accuracy of 95.03% for predicting action types and 96.42% for predicting ASD presence, demonstrating over a 10% improvement compared to models trained on single data modalities. These findings highlight the advantages of integrating multiple data modalities within the Multimodal Transformer framework.

Reinforcement Learning from Verifiable Rewards (RLVR) has recently shown that large language models (LLMs) can develop their own reasoning without direct supervision. However, applications in the medical domain, specifically for question answering, are susceptible to significant reward hacking during the reasoning phase. Our work addresses two primary forms of this behavior: i) providing a final answer without preceding reasoning, and ii) employing non-standard reasoning formats to exploit the reward mechanism. To mitigate these, we introduce a composite reward function with specific penalties for these behaviors. Our experiments show that extending RLVR with our proposed reward model leads to better-formatted reasoning with less reward hacking and good accuracy compared to the baselines. This approach marks a step toward reducing reward hacking and enhancing the reliability of models utilizing RLVR.

We discuss the appearance of superconducting strings in $E_6$ grand unification, keeping track of the magnetic monopole flux that precedes the formation of the string flux tube. This flux matching ensures compatibility with the quantum tunneling of a monopole-antimonopole pair on a metastable string. We identify two realistic $E_6$ models with superconducting (metastable) strings that also carry zero modes of the right handed Majorana neutrinos and dark matter particles. Depending on the symmetry breaking scale associated with the strings, the latter could be a source of observable gravitational waves, intermediate scale dark matter, and the observed baryon asymmetry via leptogenesis. Topologically stable superconducting strings also appear if the $E_6$ symmetry breaking leaves unbroken the $Z_2$ subgroup of $Z_4$, the center of $SO(10)$. The zero modes of the SM fermions are the charge carriers in this case. Finally, the flux matching condition ensures that the Aharanov-Bohm phase change in going around the metastable strings is an integer multiple of $2 \pi$ for all fields. The fields in the spinorial representation of SO(10) acquire a phase change of $\exp(\pm i\pi)$ if taken around the topologically stable $Z_2$ string.

We present a new scale decomposition method to investigate turbulence in wavenumber-frequency space. Using 3D magnetohydrodynamic turbulence simulations, we show that magnetic fluctuations with time scales longer than the nonlinear time exhibit an inverse cascade toward even smaller frequencies. Low frequency magnetic fluctuations support turbulence, acting as an energy reservoir that is converted into plasma kinetic energy, the latter cascading toward large wavenumbers and frequencies, where it is dissipated. Our results shed new light on the spatio-temporal properties of turbulence, potentially explaining the origin and role of low frequency turbulent fluctuations in the solar wind.

The emerging 4D millimeter-wave radar, measuring the range, azimuth, elevation, and Doppler velocity of objects, is recognized for its cost-effectiveness and robustness in autonomous driving. Nevertheless, its point clouds exhibit significant sparsity and noise, restricting its standalone application in 3D object detection. Recent 4D radar-camera fusion methods have provided effective perception. Most existing approaches, however, adopt explicit Bird's-Eye-View fusion paradigms originally designed for LiDAR-camera fusion, neglecting radar's inherent drawbacks. Specifically, they overlook the sparse and incomplete geometry of radar point clouds and restrict fusion to coarse scene-level integration. To address these problems, we propose MLF-4DRCNet, a novel two-stage framework for 3D object detection via multi-level fusion of 4D radar and camera images. Our model incorporates the point-, scene-, and proposal-level multi-modal information, enabling comprehensive feature representation. It comprises three crucial components: the Enhanced Radar Point Encoder (ERPE) module, the Hierarchical Scene Fusion Pooling (HSFP) module, and the Proposal-Level Fusion Enhancement (PLFE) module. Operating at the point-level, ERPE densities radar point clouds with 2D image instances and encodes them into voxels via the proposed Triple-Attention Voxel Feature Encoder. HSFP dynamically integrates multi-scale voxel features with 2D image features using deformable attention to capture scene context and adopts pooling to the fused features. PLFE refines region proposals by fusing image features, and further integrates with the pooled features from HSFP. Experimental results on the View-of-Delft (VoD) and TJ4DRadSet datasets demonstrate that MLF-4DRCNet achieves the state-of-the-art performance. Notably, it attains performance comparable to LiDAR-based models on the VoD dataset.

In this paper we present a consistent and distributed state estimator for multi-robot cooperative localization (CL) which efficiently fuses environmental features and loop-closure constraints across time and robots. In particular, we leverage covariance intersection (CI) to allow each robot to only estimate its own state and autocovariance and compensate for the unknown correlations between robots. Two novel multi-robot methods for utilizing common environmental SLAM features are introduced and evaluated in terms of accuracy and efficiency. Moreover, we adapt CI to enable drift-free estimation through the use of loop-closure measurement constraints to other robots' historical poses without a significant increase in computational cost. The proposed distributed CL estimator is validated against its non-realtime centralized counterpart extensively in both simulations and real-world experiments.

While dissipation in collisional plasma is defined in terms of viscosity and resistivity, the exact functional form of dissipation i.e., the so-called dissipation function in nearly collisionless plasma is unknown. Nevertheless, previous studies have suggested that there exists viscous-like energy conversion in collisionless plasma with scaling characteristics analogous to collisional plasma, and in particular that the average dissipation is proportional to the square of the rate of strain as in hydrodynamics. In this study, using 2.5D kinetic particle-in-cell (PIC) simulation of collisionless plasma turbulence, we provide an estimate of effective viscosity at each scale, obtained via a scale-filtering approach. We then compare the turbulent dynamics of the PIC simulation with that from MHD and two-fluid simulations in which with the viscosity is equal to the effective viscosity estimate obtained from the PIC simulation. We find that the global behavior in these MHD and two-fluid simulations has a striking similarity with that in its kinetic/PIC counterpart. In addition, we explore the scale dependence of the effective viscosity, and discuss implications of this approach for space plasmas.

This paper extends our previous study of gyro-emission by energetic electrons in the magnetospheres of rapidly rotating, magnetic massive stars, through a quantitative analysis of the role of Coulomb collisions with thermal electrons from stellar wind material trapped within the centrifugal magnetosphere (CM). For a dipolar field with aligned magnetic and rotational axes, we show that both gyro-cooling along magnetic loops and Coulomb cooling in the CM layer have nearly the same dependence on the magnitude and radial variation of magnetic field, implying that their ratio is a global parameter that is largely independent of the field. Analytic analysis shows that, for electrons introduced near the CM layer around a magnetic loop apex, collisional cooling is more important for electrons with high pitch angle, while more field-aligned electrons cool by gyro-emission near their mirror point close to the loop base. Numerical models that assume a gyrotropic initial deposition with a gaussian distribution in both radius and loop co-latitude show the residual gyro-emission is generally strongest near the loop base, with highly relativistic electrons suffering much lower collisional losses than lower-energy electrons that are only mildly relativistic. Finally, we briefly discuss the potential applicability of this formalism to magnetic ultracool dwarfs, for which VLBI observations indicate incoherent radio emission to be concentrated around the magnetic equator, in contrast to our predictions here for magnetic hot stars. We suggest that this difference could be attributed to either a lower ambient density of thermal electrons, or more highly relativistic non-thermal electrons, both of which would reduce the relative importance of the collisional cooling explored here.

Knowledge graphs, collectively as a knowledge network, have become critical tools for knowledge discovery in computable and explainable knowledge systems. Due to the semantic and structural complexities of biomedical data, these knowledge graphs need to enable dynamic reasoning over large evolving graphs and support fit-for-purpose abstraction, while establishing standards, preserving provenance and enforcing policy constraints for actionable discovery. A recent meeting of leading scientists discussed the opportunities, challenges and future directions of a biomedical knowledge network. Here we present six desiderata inspired by the meeting: (1) inference and reasoning in biomedical knowledge graphs need domain-centric approaches; (2) harmonized and accessible standards are required for knowledge graph representation and metadata; (3) robust validation of biomedical knowledge graphs needs multi-layered, context-aware approaches that are both rigorous and scalable; (4) the evolving and synergistic relationship between knowledge graphs and large language models is essential in empowering AI-driven biomedical discovery; (5) integrated development environments, public repositories, and governance frameworks are essential for secure and reproducible knowledge graph sharing; and (6) robust validation, provenance, and ethical governance are critical for trustworthy biomedical knowledge graphs. Addressing these key issues will be essential to realize the promises of a biomedical knowledge network in advancing biomedicine.

This paper presents the submission of the UDInfo team to the SIGIR 2025 LiveRAG Challenge. We introduce PreQRAG, a Retrieval Augmented Generation (RAG) architecture designed to improve retrieval and generation quality through targeted question preprocessing. PreQRAG incorporates a pipeline that first classifies each input question as either single-document or multi-document type. For single-document questions, we employ question rewriting techniques to improve retrieval precision and generation relevance. For multi-document questions, we decompose complex queries into focused sub-questions that can be processed more effectively by downstream components. This classification and rewriting strategy improves the RAG performance. Experimental evaluation of the LiveRAG Challenge dataset demonstrates the effectiveness of our question-type-aware architecture, with PreQRAG achieving the preliminary second place in Session 2 of the LiveRAG challenge.

Proteins are fundamental to biology, executing diverse functions through complex physicochemical interactions, and they hold transformative potential across medicine, materials science, and environmental applications. Protein Language Models (pLMs) aim to unlock insights from the vast space of unlabeled protein sequences by learning rich, semantic representations from primary sequences via masked language modeling. However, these models typically exhibit limited generative capacity. In this work, we introduce the Diffusion Sequence Model (DSM), a novel pLM trained with masked diffusion to enable both high-quality representation learning and generative protein design. DSM builds upon the ESM2 architecture by incorporating a masked forward diffusion process inspired by the LLaDA framework. After training, DSM is capable of generating diverse, biomimetic sequences that align with expected amino acid compositions, secondary structures, and predicted functions, even with 90\% token corruption. Furthermore, DSM's learned representations match or exceed those of similarly sized pLMs on downstream tasks. We also introduce DSM(ppi), a variant fine-tuned to generate protein binders by attending to target sequences. We demonstrate DSM(ppi)'s effectiveness on the challenging Bench-tested Binder Benchmark (BenchBB), where both DSM and DSM(ppi) produce candidates with superior predicted binding affinity compared to known binders. Our results establish masked diffusion as a powerful paradigm for unifying protein representation and generation in a single framework.