Search

Offline preference-based reinforcement learning (PbRL) typically operates in two phases: first, use human preferences to learn a reward model and annotate rewards for a reward-free offline dataset; second, learn a policy by optimizing the learned reward via offline RL. However, accurately modeling step-wise rewards from trajectory-level preference feedback presents inherent challenges. The reward bias introduced, particularly the overestimation of predicted rewards, leads to optimistic trajectory stitching, which undermines the pessimism mechanism critical to the offline RL phase. To address this challenge, we propose In-Dataset Trajectory Return Regularization (DTR) for offline PbRL, which leverages conditional sequence modeling to mitigate the risk of learning inaccurate trajectory stitching under reward bias. Specifically, DTR employs Decision Transformer and TD-Learning to strike a balance between maintaining fidelity to the behavior policy with high in-dataset trajectory returns and selecting optimal actions based on high reward labels. Additionally, we introduce an ensemble normalization technique that effectively integrates multiple reward models, balancing the tradeoff between reward differentiation and accuracy. Empirical evaluations on various benchmarks demonstrate the superiority of DTR over other state-of-the-art baselines, Comment: 7 pages, Proceedings of the 39th AAAI Conference on Artificial Intelligence (AAAI-25)

Large Language models (LLMs) have become a research hotspot. To accelerate the inference of LLMs, storing computed caches in memory has become the standard technique. However, as the inference length increases, growing KV caches might lead to out-of-memory issues. Many existing methods address this issue through KV cache compression, primarily by preserving key tokens throughout all layers to reduce information loss. Most of them allocate a uniform budget size for each layer to retain. However, we observe that the minimum budget sizes needed to retain essential information vary across layers and models based on the perspectives of attention and hidden state output. Building on this observation, this paper proposes a simple yet effective KV cache compression method that leverages layer uncertainty to allocate budget size for each layer. Experimental results show that the proposed method can reduce memory usage of the KV caches to only $\sim$20\% when compared to Full KV inference while achieving nearly lossless performance.

Organic Solar Cells (OSCs) are a promising technology for sustainable energy production. However, the identification of molecules with desired OSC properties typically involves laborious experimental research. To accelerate progress in the field, it is crucial to develop machine learning models capable of accurately predicting the properties of OSC molecules. While graph representation learning has demonstrated success in molecular property prediction, it remains underexplored for OSC-specific tasks. Existing methods fail to capture the unique structural features of OSC molecules, particularly the intricate ring systems that critically influence OSC properties, leading to suboptimal performance. To fill the gap, we present RingFormer, a novel graph transformer framework specially designed to capture both atom and ring level structural patterns in OSC molecules. RingFormer constructs a hierarchical graph that integrates atomic and ring structures and employs a combination of local message passing and global attention mechanisms to generate expressive graph representations for accurate OSC property prediction. We evaluate RingFormer's effectiveness on five curated OSC molecule datasets through extensive experiments. The results demonstrate that RingFormer consistently outperforms existing methods, achieving a 22.77% relative improvement over the nearest competitor on the CEPDB dataset., Comment: 12 pages, 4 figures. This is the extended version of the paper accepted at AAAI 2025, which includes all technical appendices and additional experimental details

Deep learning has transformed AI applications but faces critical security challenges, including adversarial attacks, data poisoning, model theft, and privacy leakage. This survey examines these vulnerabilities, detailing their mechanisms and impact on model integrity and confidentiality. Practical implementations, including adversarial examples, label flipping, and backdoor attacks, are explored alongside defenses such as adversarial training, differential privacy, and federated learning, highlighting their strengths and limitations. Advanced methods like contrastive and self-supervised learning are presented for enhancing robustness. The survey concludes with future directions, emphasizing automated defenses, zero-trust architectures, and the security challenges of large AI models. A balanced approach to performance and security is essential for developing reliable deep learning systems.

The prevalence of unhealthy eating habits has become an increasingly concerning issue in the United States. However, major food recommendation platforms (e.g., Yelp) continue to prioritize users' dietary preferences over the healthiness of their choices. Although efforts have been made to develop health-aware food recommendation systems, the personalization of such systems based on users' specific health conditions remains under-explored. In addition, few research focus on the interpretability of these systems, which hinders users from assessing the reliability of recommendations and impedes the practical deployment of these systems. In response to this gap, we first establish two large-scale personalized health-aware food recommendation benchmarks at the first attempt. We then develop a novel framework, Multi-Objective Personalized Interpretable Health-aware Food Recommendation System (MOPI-HFRS), which provides food recommendations by jointly optimizing the three objectives: user preference, personalized healthiness and nutritional diversity, along with an large language model (LLM)-enhanced reasoning module to promote healthy dietary knowledge through the interpretation of recommended results. Specifically, this holistic graph learning framework first utilizes two structure learning and a structure pooling modules to leverage both descriptive features and health data. Then it employs Pareto optimization to achieve designed multi-facet objectives. Finally, to further promote the healthy dietary knowledge and awareness, we exploit an LLM by utilizing knowledge-infusion, prompting the LLMs with knowledge obtained from the recommendation model for interpretation.

Fairness in multi-document summarization (MDS) measures whether a system can generate a summary fairly representing information from documents with different social attribute values. Fairness in MDS is crucial since a fair summary can offer readers a comprehensive view. Previous works focus on quantifying summary-level fairness using Proportional Representation, a fairness measure based on Statistical Parity. However, Proportional Representation does not consider redundancy in input documents and overlooks corpus-level unfairness. In this work, we propose a new summary-level fairness measure, Equal Coverage, which is based on coverage of documents with different social attribute values and considers the redundancy within documents. To detect the corpus-level unfairness, we propose a new corpus-level measure, Coverage Parity. Our human evaluations show that our measures align more with our definition of fairness. Using our measures, we evaluate the fairness of thirteen different LLMs. We find that Claude3-sonnet is the fairest among all evaluated LLMs. We also find that almost all LLMs overrepresent different social attribute values.

A complete understanding of the initial conditions of high-mass star formation and what processes determine multiplicity require the study of the magnetic field (B-field) in young, massive cores. Using ALMA 250 GHz polarization (0.3" = 1000 au) and ALMA 220 GHz high-angular resolution observations (0.05" = 160 au), we have performed a full energy analysis including the B-field at core scales and have assessed what influences the multiplicity inside a massive core previously believed to be in the prestellar phase. With 31 Msun, the G11.92 MM2 core has a young CS outflow with a dynamical time scale of a few thousand years. At high-resolution, the MM2 core fragments into a binary system with a projected separation of 505 au and a binary mass ratio of 1.14. Using the DCF method with an ADF analysis, we estimate in this core a B-field strength of 6.2 mG and a mass-to-flux ratio of 18. The MM2 core is strongly subvirialized with a virial parameter of 0.064, including the B-field. The high mass-to-flux ratio and low virial parameter indicate that this massive core is very likely undergoing runaway collapse, which is in direct contradiction with the core-accretion model. The MM2 core is embedded in a filament that has a velocity gradient consistent with infall. In line with clump-fed scenarios, the core can grow in mass at a rate of 1.9--5.6 x 10^-4 Msun/yr. In spite of the B-field having only a minor contribution to the total energy budget at core scales, it likely plays a more important role at smaller scales by setting the binary properties. Considering energy ratios and a fragmentation criterion at the core scale, the binary could have been formed by core fragmentation. The binary properties (separation and mass ratio), however, are also consistent with radiation-magnetohydrodynamic simulations with super-Alfvenic, supersonic (or sonic) turbulence that form binaries by disk fragmentation., Comment: Accepted for publications in ApJ (9 pages, 3 figures, Appendix)

Modeling the evolutions of driving scenarios is important for the evaluation and decision-making of autonomous driving systems. Most existing methods focus on one aspect of scene evolution such as map generation, motion prediction, and trajectory planning. In this paper, we propose a unified Generative Pre-training for Driving (GPD-1) model to accomplish all these tasks altogether without additional fine-tuning. We represent each scene with ego, agent, and map tokens and formulate autonomous driving as a unified token generation problem. We adopt the autoregressive transformer architecture and use a scene-level attention mask to enable intra-scene bi-directional interactions. For the ego and agent tokens, we propose a hierarchical positional tokenizer to effectively encode both 2D positions and headings. For the map tokens, we train a map vector-quantized autoencoder to efficiently compress ego-centric semantic maps into discrete tokens. We pre-train our GPD-1 on the large-scale nuPlan dataset and conduct extensive experiments to evaluate its effectiveness. With different prompts, our GPD-1 successfully generalizes to various tasks without finetuning, including scene generation, traffic simulation, closed-loop simulation, map prediction, and motion planning. Code: https://github.com/wzzheng/GPD., Comment: Code is available at: https://github.com/wzzheng/GPD

Current collaborative perception methods often rely on fully annotated datasets, which can be expensive to obtain in practical situations. To reduce annotation costs, some works adopt sparsely supervised learning techniques and generate pseudo labels for the missing instances. However, these methods fail to achieve an optimal confidence threshold that harmonizes the quality and quantity of pseudo labels. To address this issue, we propose an end-to-end Collaborative perception Dual Teacher-Student framework (CoDTS), which employs adaptive complementary learning to produce both high-quality and high-quantity pseudo labels. Specifically, the Main Foreground Mining (MFM) module generates high-quality pseudo labels based on the prediction of the static teacher. Subsequently, the Supplement Foreground Mining (SFM) module ensures a balance between the quality and quantity of pseudo labels by adaptively identifying missing instances based on the prediction of the dynamic teacher. Additionally, the Neighbor Anchor Sampling (NAS) module is incorporated to enhance the representation of pseudo labels. To promote the adaptive complementary learning, we implement a staged training strategy that trains the student and dynamic teacher in a mutually beneficial manner. Extensive experiments demonstrate that the CoDTS effectively ensures an optimal balance of pseudo labels in both quality and quantity, establishing a new state-of-the-art in sparsely supervised collaborative perception., Comment: AAAI 2025

Parameter-efficient transfer learning (PETL) has become a promising paradigm for adapting large-scale vision foundation models to downstream tasks. Typical methods primarily leverage the intrinsic low rank property to make decomposition, learning task-specific weights while compressing parameter size. However, such approaches predominantly manipulate within the original feature space utilizing a single-branch structure, which might be suboptimal for decoupling the learned representations and patterns. In this paper, we propose ALoRE, a novel PETL method that reuses the hypercomplex parameterized space constructed by Kronecker product to Aggregate Low Rank Experts using a multi-branch paradigm, disentangling the learned cognitive patterns during training. Thanks to the artful design, ALoRE maintains negligible extra parameters and can be effortlessly merged into the frozen backbone via re-parameterization in a sequential manner, avoiding additional inference latency. We conduct extensive experiments on 24 image classification tasks using various backbone variants. Experimental results demonstrate that ALoRE outperforms the full fine-tuning strategy and other state-of-the-art PETL methods in terms of performance and parameter efficiency. For instance, ALoRE obtains 3.06% and 9.97% Top-1 accuracy improvement on average compared to full fine-tuning on the FGVC datasets and VTAB-1k benchmark by only updating 0.15M parameters., Comment: 23 pages, 7 figures

As a common form of communication in social media,stickers win users' love in the internet scenarios, for their ability to convey emotions in a vivid, cute, and interesting way. People prefer to get an appropriate sticker through retrieval rather than creation for the reason that creating a sticker is time-consuming and relies on rule-based creative tools with limited capabilities. Nowadays, advanced text-to-video algorithms have spawned numerous general video generation systems that allow users to customize high-quality, photo-realistic videos by only providing simple text prompts. However, creating customized animated stickers, which have lower frame rates and more abstract semantics than videos, is greatly hindered by difficulties in data acquisition and incomplete benchmarks. To facilitate the exploration of researchers in animated sticker generation (ASG) field, we firstly construct the currently largest vision-language sticker dataset named VSD2M at a two-million scale that contains static and animated stickers. Secondly, to improve the performance of traditional video generation methods on ASG tasks with discrete characteristics, we propose a Spatial Temporal Interaction (STI) layer that utilizes semantic interaction and detail preservation to address the issue of insufficient information utilization. Moreover, we train baselines with several video generation methods (e.g., transformer-based, diffusion-based methods) on VSD2M and conduct a detailed analysis to establish systemic supervision on ASG task. To the best of our knowledge, this is the most comprehensive large-scale benchmark for multi-frame animated sticker generation, and we hope this work can provide valuable inspiration for other scholars in intelligent creation., Comment: 11 pages, 9 figures

Proline (Pro) is one kind of proteinogenic amino acid and an important signaling molecule in the process of metabolism. Hydroxyproline (Hyp) is a product on Pro oxygen sensing post-translational modification (PTM), which is efficiently modulated tumor cells for angiogenesis. Distinguishing between Pro and Hyp is crucial for diagnosing connective tissue disorders, as elevated levels of Hyp can indicate abnormal collagen metabolism, often associated with diseases like osteogenesis imperfecta or fibrosis. However, there is a very small difference between molecular structures of Pro and Hyp, which is a big challenge for current detection technologies to distinguish them. For surface-enhanced Raman scattering (SERS) sensors, the similar molecule structure leads to similar Raman spectra that are difficult to distinguish. Furthermore, another problem is the weak affinity between amino acids sample and SERS-active substrates by physical adsorption. The selecting capturing of Pro and Hyp in the mixture of amino acids is not easy to achieve. In this work, we designed a new method for Pro and Hyp specifical detection and recognition by using gold nanopillars as the SERS substrate and combing nitrilotriacetic acid (NTA) with nickel (Ni) to form NTA-Ni structure as a specifical affinity agent. One side of NTA-Ni was attached to gold nanopillars through thiol binding. Another side captured the amino acids using reversible binding by receptor-ligand interaction between Ni and amino acids. Because of the different binding time with NTA-Ni and amino acids, the sensor can recognize Pro and Hyp from amino acids mixture. Then we used automatic peak assignment program for data analysis and machine learning model to distinguish between Pro and Hyp. The label-free SERS detection of amino acids PTM using gold nanopillars provides a potential method to further biomolecule detection and specifical capture.

It is well known that LLM-based systems are data-hungry. Recent LLM-based TTS works typically employ complex data processing pipelines to obtain high-quality training data. These sophisticated pipelines require excellent models at each stage (e.g., speech denoising, speech enhancement, speaker diarization, and punctuation models), which themselves demand high-quality training data and are rarely open-sourced. Even with state-of-the-art models, issues persist, such as incomplete background noise removal and misalignment between punctuation and actual speech pauses. Moreover, the stringent filtering strategies often retain only 10-30\% of the original data, significantly impeding data scaling efforts. In this work, we leverage a noise-robust audio tokenizer (S3Tokenizer) to design a simplified yet effective TTS data processing pipeline that maintains data quality while substantially reducing data acquisition costs, achieving a data retention rate of over 50\%. Beyond data scaling challenges, LLM-based TTS systems also incur higher deployment costs compared to conventional approaches. Current systems typically use LLMs solely for text-to-token generation, while requiring separate models (e.g., flow matching models) for token-to-waveform generation, which cannot be directly executed by LLM inference engines, further complicating deployment. To address these challenges, we eliminate redundant modules in both LLM and flow components, replacing the flow model backbone with an LLM architecture. Building upon this simplified flow backbone, we propose a unified architecture for both streaming and non-streaming inference, significantly reducing deployment costs. Finally, we explore the feasibility of unifying TTS and ASR tasks using the same data for training, thanks to the simplified pipeline and the S3Tokenizer that reduces the quality requirements for TTS training data., Comment: Technical Report

Due to the scarcity and unpredictable nature of defect samples, industrial anomaly detection (IAD) predominantly employs unsupervised learning. However, all unsupervised IAD methods face a common challenge: the inherent bias in normal samples, which causes models to focus on variable regions while overlooking potential defects in invariant areas. To effectively overcome this, it is essential to decompose and recalibrate attention, guiding the model to suppress irrelevant variations and concentrate on subtle, defect-susceptible areas. In this paper, we propose Recalibrating Attention of Industrial Anomaly Detection (RAAD), a framework that systematically decomposes and recalibrates attention maps. RAAD employs a two-stage process: first, it reduces attention bias through quantization, and second, it fine-tunes defect-prone regions for improved sensitivity. Central to this framework is Hierarchical Quantization Scoring (HQS), which dynamically allocates bit-widths across layers based on their anomaly detection contributions. HQS dynamically adjusts bit-widths based on the hierarchical nature of attention maps, compressing lower layers that produce coarse and noisy attention while preserving deeper layers with sharper, defect-focused attention. This approach optimizes both computational efficiency and the model' s sensitivity to anomalies. We validate the effectiveness of RAAD on 32 datasets using a single 3090ti. Experiments demonstrate that RAAD, balances the complexity and expressive power of the model, enhancing its anomaly detection capability.

Photonic neural networks (PNNs) have garnered significant interest due to their potential to offer low latency, high bandwidth, and energy efficiency in neuromorphic computing and machine learning. In PNNs, weights are photonic devices that make them susceptible to environmental factors and fabrication variations. These vulnerabilities can result in inaccurate parameter mapping, increased tuning power consumption, and reduced network performance when conventional offline training methods are used. Here, we experimentally demonstrate an online training and pruning method to address these challenges. By incorporating a power-related term into the conventional loss function, our approach minimizes the inference power budget. With this method, PNNs achieve 96% accuracy while reducing the power consumption by almost 45% on the Iris dataset, despite fabrication and thermal variations. Furthermore, our method is validated with a two-layer convolutional neural network (CNN) experiment for radio-frequency (RF) fingerprinting applications and simulations across larger CNNs on image classification datasets, including MNIST, CIFAR-10, and CIFAR-100. This work represents a significant milestone in enabling adaptive online training of PNNs and showcases their potential for real-world applications., Comment: 16 pages, 6 figures

Image inpainting is an important image generation task, which aims to restore corrupted image from partial visible area. Recently, diffusion Schr\"odinger bridge methods effectively tackle this task by modeling the translation between corrupted and target images as a diffusion Schr\"odinger bridge process along a noising schedule path. Although these methods have shown superior performance, in this paper, we find that 1) existing methods suffer from a schedule-restoration mismatching issue, i.e., the theoretical schedule and practical restoration processes usually exist a large discrepancy, which theoretically results in the schedule not fully leveraged for restoring images; and 2) the key reason causing such issue is that the restoration process of all pixels are actually asynchronous but existing methods set a synchronous noise schedule to them, i.e., all pixels shares the same noise schedule. To this end, we propose a schedule-Asynchronous Diffusion Schr\"odinger Bridge (AsyncDSB) for image inpainting. Our insight is preferentially scheduling pixels with high frequency (i.e., large gradients) and then low frequency (i.e., small gradients). Based on this insight, given a corrupted image, we first train a network to predict its gradient map in corrupted area. Then, we regard the predicted image gradient as prior and design a simple yet effective pixel-asynchronous noise schedule strategy to enhance the diffusion Schr\"odinger bridge. Thanks to the asynchronous schedule at pixels, the temporal interdependence of restoration process between pixels can be fully characterized for high-quality image inpainting. Experiments on real-world datasets show that our AsyncDSB achieves superior performance, especially on FID with around 3% - 14% improvement over state-of-the-art baseline methods., Comment: Accepted by AAAI 2025

Event cameras, as bio-inspired sensors, are asynchronously triggered with high-temporal resolution compared to intensity cameras. Recent work has focused on fusing the event measurements with inertial measurements to enable ego-motion estimation in high-speed and HDR environments. However, existing methods predominantly rely on IMU preintegration designed mainly for synchronous sensors and discrete-time frameworks. In this paper, we propose a continuous-time preintegration method based on the Temporal Gaussian Process (TGP) called GPO. Concretely, we model the preintegration as a time-indexed motion trajectory and leverage an efficient two-step optimization to initialize the precision preintegration pseudo-measurements. Our method realizes a linear and constant time cost for initialization and query, respectively. To further validate the proposal, we leverage the GPO to design an asynchronous event-inertial odometry and compare with other asynchronous fusion schemes within the same odometry system. Experiments conducted on both public and own-collected datasets demonstrate that the proposed GPO offers significant advantages in terms of precision and efficiency, outperforming existing approaches in handling asynchronous sensor fusion., Comment: 8pages

We present phi-4, a 14-billion parameter language model developed with a training recipe that is centrally focused on data quality. Unlike most language models, where pre-training is based primarily on organic data sources such as web content or code, phi-4 strategically incorporates synthetic data throughout the training process. While previous models in the Phi family largely distill the capabilities of a teacher model (specifically GPT-4), phi-4 substantially surpasses its teacher model on STEM-focused QA capabilities, giving evidence that our data-generation and post-training techniques go beyond distillation. Despite minimal changes to the phi-3 architecture, phi-4 achieves strong performance relative to its size -- especially on reasoning-focused benchmarks -- due to improved data, training curriculum, and innovations in the post-training scheme.

Blue Large-Amplitude Pulsators (BLAPs) represent a recently identified class of pulsating stars distinguished by their short pulsation periods (22-40 minutes) and asymmetric light curves. This study investigates the evolutionary channel of HD 133729, the first confirmed BLAP in a binary system. Through binary evolution simulations with MESA, we explored various mass ratios and initial orbital periods, finding that a mass ratio of q = 0.30 coupled with non-conservative mass transfer ($\rm \beta$ = 0.15) successfully reproduces the observational characteristics (including luminosity, surface gravity, and effective temperature) of the binary system. Our models not only match the observed pulsational properties but also predict significant helium and nitrogen enhancements on the surface of the main-sequence companion. The system is expected to eventually undergo a common envelope phase leading to a stellar merger. Our findings provide crucial insights into the formation mechanism and evolutionary fate of BLAPs with main-sequence companions, while also placing constraints on the elemental abundances of their binary companions., Comment: Comments are welcome

Automated drug discovery offers significant potential for accelerating the development of novel therapeutics by substituting labor-intensive human workflows with machine-driven processes. However, a critical bottleneck persists in the inability of current automated frameworks to assess whether newly designed molecules infringe upon existing patents, posing significant legal and financial risks. We introduce PatentFinder, a novel tool-enhanced and multi-agent framework that accurately and comprehensively evaluates small molecules for patent infringement. It incorporates both heuristic and model-based tools tailored for decomposed subtasks, featuring: MarkushParser, which is capable of optical chemical structure recognition of molecular and Markush structures, and MarkushMatcher, which enhances large language models' ability to extract substituent groups from molecules accurately. On our benchmark dataset MolPatent-240, PatentFinder outperforms baseline approaches that rely solely on large language models, demonstrating a 13.8\% increase in F1-score and a 12\% rise in accuracy. Experimental results demonstrate that PatentFinder mitigates label bias to produce balanced predictions and autonomously generates detailed, interpretable patent infringement reports. This work not only addresses a pivotal challenge in automated drug discovery but also demonstrates the potential of decomposing complex scientific tasks into manageable subtasks for specialized, tool-augmented agents.

Current video diffusion models achieve impressive generation quality but struggle in interactive applications due to bidirectional attention dependencies. The generation of a single frame requires the model to process the entire sequence, including the future. We address this limitation by adapting a pretrained bidirectional diffusion transformer to a causal transformer that generates frames on-the-fly. To further reduce latency, we extend distribution matching distillation (DMD) to videos, distilling 50-step diffusion model into a 4-step generator. To enable stable and high-quality distillation, we introduce a student initialization scheme based on teacher's ODE trajectories, as well as an asymmetric distillation strategy that supervises a causal student model with a bidirectional teacher. This approach effectively mitigates error accumulation in autoregressive generation, allowing long-duration video synthesis despite training on short clips. Our model supports fast streaming generation of high quality videos at 9.4 FPS on a single GPU thanks to KV caching. Our approach also enables streaming video-to-video translation, image-to-video, and dynamic prompting in a zero-shot manner. We will release the code based on an open-source model in the future., Comment: Project Page: https://causvid.github.io/

We introduce UniReal, a unified framework designed to address various image generation and editing tasks. Existing solutions often vary by tasks, yet share fundamental principles: preserving consistency between inputs and outputs while capturing visual variations. Inspired by recent video generation models that effectively balance consistency and variation across frames, we propose a unifying approach that treats image-level tasks as discontinuous video generation. Specifically, we treat varying numbers of input and output images as frames, enabling seamless support for tasks such as image generation, editing, customization, composition, etc. Although designed for image-level tasks, we leverage videos as a scalable source for universal supervision. UniReal learns world dynamics from large-scale videos, demonstrating advanced capability in handling shadows, reflections, pose variation, and object interaction, while also exhibiting emergent capability for novel applications., Comment: webpage: https://xavierchen34.github.io/UniReal-Page/

Document content extraction is crucial in computer vision, especially for meeting the high-quality data needs of large language models (LLMs) and retrieval-augmented generation (RAG) technologies. However, current document parsing methods suffer from significant limitations in terms of diversity and comprehensive evaluation. To address these challenges, we introduce OmniDocBench, a novel multi-source benchmark designed to advance automated document content extraction. OmniDocBench includes a meticulously curated and annotated high-quality evaluation dataset comprising nine diverse document types, such as academic papers, textbooks, slides, among others. Our benchmark provides a flexible and comprehensive evaluation framework with 19 layout category labels and 14 attribute labels, enabling multi-level assessments across entire datasets, individual modules, or specific data types. Using OmniDocBench, we perform an exhaustive comparative analysis of existing modular pipelines and multimodal end-to-end methods, highlighting their limitations in handling document diversity and ensuring fair evaluation. OmniDocBench establishes a robust, diverse, and fair evaluation standard for the document content extraction field, offering crucial insights for future advancements and fostering the development of document parsing technologies. The codes and dataset is available in https://github.com/opendatalab/OmniDocBench.

LiDAR is widely used in Simultaneous Localization and Mapping (SLAM) and autonomous driving. The LiDAR odometry is of great importance in multi-sensor fusion. However, in some unstructured environments, the point cloud registration cannot constrain the poses of the LiDAR due to its sparse geometric features, which leads to the degeneracy of multi-sensor fusion accuracy. To address this problem, we propose a novel real-time approach to sense and compensate for the degeneracy of LiDAR. Firstly, this paper introduces the degeneracy factor with clear meaning, which can measure the degeneracy of LiDAR. Then, the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering method adaptively perceives the degeneracy with better environmental generalization. Finally, the degeneracy perception results are utilized to fuse LiDAR and IMU, thus effectively resisting degeneracy effects. Experiments on our dataset show the method's high accuracy and robustness and validate our algorithm's adaptability to different environments and LiDAR scanning modalities., Comment: 13 pages

In few-shot action recognition~(FSAR), long sub-sequences of video naturally express entire actions more effectively. However, the computational complexity of mainstream Transformer-based methods limits their application. Recent Mamba demonstrates efficiency in modeling long sequences, but directly applying Mamba to FSAR overlooks the importance of local feature modeling and alignment. Moreover, long sub-sequences within the same class accumulate intra-class variance, which adversely impacts FSAR performance. To solve these challenges, we propose a \underline{\textbf{M}}atryoshka M\underline{\textbf{A}}mba and Co\underline{\textbf{N}}tras\underline{\textbf{T}}ive Le\underline{\textbf{A}}rning framework~(\textbf{Manta}). Firstly, the Matryoshka Mamba introduces multiple Inner Modules to enhance local feature representation, rather than directly modeling global features. An Outer Module captures dependencies of timeline between these local features for implicit temporal alignment. Secondly, a hybrid contrastive learning paradigm, combining both supervised and unsupervised methods, is designed to mitigate the negative effects of intra-class variance accumulation. The Matryoshka Mamba and the hybrid contrastive learning paradigm operate in parallel branches within Manta, enhancing Mamba for FSAR of long sub-sequence. Manta achieves new state-of-the-art performance on prominent benchmarks, including SSv2, Kinetics, UCF101, and HMDB51. Extensive empirical studies prove that Manta significantly improves FSAR of long sub-sequence from multiple perspectives. The code is released at https://github.com/wenbohuang1002/Manta., Comment: Accepted by AAAI 2025

Recent advances in embodied agents with multimodal perception and reasoning capabilities based on large vision-language models (LVLMs), excel in autonomously interacting either real or cyber worlds, helping people make intelligent decisions in complex environments. However, the current works are normally optimized by golden action trajectories or ideal task-oriented solutions toward a definitive goal. This paradigm considers limited user-oriented factors, which could be the reason for their performance reduction in a wide range of personal assistant applications. To address this, we propose Chain-of-User-Thought (COUT), a novel embodied reasoning paradigm that takes a chain of thought from basic action thinking to explicit and implicit personalized preference thought to incorporate personalized factors into autonomous agent learning. To target COUT, we introduce SmartAgent, an agent framework perceiving cyber environments and reasoning personalized requirements as 1) interacting with GUI to access an item pool, 2) generating users' explicit requirements implied by previous actions, and 3) recommending items to fulfill users' implicit requirements. To demonstrate SmartAgent's capabilities, we also create a brand-new dataset SmartSpot that offers a full-stage personalized action-involved environment. To our best knowledge, our work is the first to formulate the COUT process, serving as a preliminary attempt towards embodied personalized agent learning. Our extensive experiments on SmartSpot illuminate SmartAgent's functionality among a series of embodied and personalized sub-tasks. We will release code and data upon paper notification at \url{https://github.com/tsinghua-fib-lab/SmartAgent}.

We introduce CADSpotting, an efficient method for panoptic symbol spotting in large-scale architectural CAD drawings. Existing approaches struggle with the diversity of symbols, scale variations, and overlapping elements in CAD designs. CADSpotting overcomes these challenges by representing each primitive with dense points instead of a single primitive point, described by essential attributes like coordinates and color. Building upon a unified 3D point cloud model for joint semantic, instance, and panoptic segmentation, CADSpotting learns robust feature representations. To enable accurate segmentation in large, complex drawings, we further propose a novel Sliding Window Aggregation (SWA) technique, combining weighted voting and Non-Maximum Suppression (NMS). Moreover, we introduce a large-scale CAD dataset named LS-CAD to support our experiments. Each floorplan in LS-CAD has an average coverage of 1,000 square meter(versus 100 square meter in the existing dataset), providing a valuable benchmark for symbol spotting research. Experimental results on FloorPlanCAD and LS-CAD datasets demonstrate that CADSpotting outperforms existing methods, showcasing its robustness and scalability for real-world CAD applications., Comment: 16pages, 12 figures, Project web-page: https://dgeneai.github.io/cadspotting-pages/

Story visualization has gained increasing attention in artificial intelligence. However, existing methods still struggle with maintaining a balance between character identity preservation and text-semantics alignment, largely due to a lack of detailed semantic modeling of the story scene. To tackle this challenge, we propose a novel knowledge graph, namely Character Graph (\textbf{CG}), which comprehensively represents various story-related knowledge, including the characters, the attributes related to characters, and the relationship between characters. We then introduce StoryWeaver, an image generator that achieve Customization via Character Graph (\textbf{C-CG}), capable of consistent story visualization with rich text semantics. To further improve the multi-character generation performance, we incorporate knowledge-enhanced spatial guidance (\textbf{KE-SG}) into StoryWeaver to precisely inject character semantics into generation. To validate the effectiveness of our proposed method, extensive experiments are conducted using a new benchmark called TBC-Bench. The experiments confirm that our StoryWeaver excels not only in creating vivid visual story plots but also in accurately conveying character identities across various scenarios with considerable storage efficiency, \emph{e.g.}, achieving an average increase of +9.03\% DINO-I and +13.44\% CLIP-T. Furthermore, ablation experiments are conducted to verify the superiority of the proposed module. Codes and datasets are released at https://github.com/Aria-Zhangjl/StoryWeaver.

Reconfigurable Intelligent Surface (RIS) is considered as a promising technology for 6G due to its ability to actively modify the electromagnetic propagation environment. Accurate channel modeling is essential for the design and evaluation of RIS assisted communication systems. Most current research models the RIS channel as a cascade of Tx-RIS and RIS-Rx sub-channels. However, most validation efforts regarding this assumption focus on large-scale path loss. To further explore this, in this paper, we derive and extend a convolution expression of RIS cascaded channel model based on the previously proposed Geometry-based Stochastic Model (GBSM)-based RIS cascaded channels. This model follows the 3GPP standard framework and leverages parameters such as angles, delays, and path powers defined in the GBSM model to more accurately reflect the smallscale characteristics of RIS multipath cascades. To verify the accuracy of this model, we conduct measurements of the TxRIS-Rx channel, Tx-RIS, and RIS-Rx sub-channels in a factory environment at 6.9 GHz, using the measured data to demonstrate the models validity and applicability in real-world scenarios. Validation with measured data shows that the proposed model accurately describes the characteristics of the RIS cascaded channel in terms of delay, angle, and power in complex multipath environments, providing important references for the design and deployment of RIS systems.

As 6G research advances, the growing demand leads to the emergence of novel technologies such as Integrated Sensing and Communication (ISAC), new antenna arrays like Extremely Large MIMO (XL-MIMO) and Reconfigurable Intelligent Surfaces (RIS), along with multi-frequency bands (6-24 GHz, above 100 GHz). Standardized unified channel models are crucial for research and performance evaluation across generations of mobile communication, but the existing 5G 3GPP channel model based on geometry-based stochastic model (GBSM) requires further extension to accommodate these 6G technologies. In response to this need, this article first investigates six distinctive channel characteristics introduced by 6G techenologies, such as ISAC target RCS, sparsity in the new mid-band, and others. Subsequently, an extended GBSM (E-GBSM) is proposed, integrating these characteristics into a unified modeling framework. The proposed model not only accommodates 6G technologies with flexibility but also maintains backward compatibility with 5G, ensuring a smooth evolution between generations. Finally, the implementation process of the proposed model is detailed, with experiments and simulations validate its effectiveness and accuracy, providing support for 6G channel modeling standardization efforts.

Continual Semantic Parsing (CSP) aims to train parsers to convert natural language questions into SQL across tasks with limited annotated examples, adapting to the real-world scenario of dynamically updated databases. Previous studies mitigate this challenge by replaying historical data or employing parameter-efficient tuning (PET), but they often violate data privacy or rely on ideal continual learning settings. To address these problems, we propose a new Large Language Model (LLM)-Enhanced Continuous Semantic Parsing method, named LECSP, which alleviates forgetting while encouraging generalization, without requiring real data replay or ideal settings. Specifically, it first analyzes the commonalities and differences between tasks from the SQL syntax perspective to guide LLMs in reconstructing key memories and improving memory accuracy through a calibration strategy. Then, it uses a task-aware dual-teacher distillation framework to promote the accumulation and transfer of knowledge during sequential training. Experimental results on two CSP benchmarks show that our method significantly outperforms existing methods, even those utilizing data replay or ideal settings. Additionally, we achieve generalization performance beyond the upper limits, better adapting to unseen tasks.

The Sine-Cosine function, which is widely adopted in mathematics and physics, has attracted our attention due to its unique properties. By delving into the coupling effect of the Sine-Cosine function, we discover a previously unreported class of nonlinear systems, namely the Sine-Cosine Nonlinear System Family (SCNSF). This discovery is motivated by the need to expand the repertoire of nonlinear systems and understand the complex behaviors that can emerge from the combination of basic trigonometric functions. The SCNSF has both chaotic characteristics in the real number domain and fractal characteristics in the complex number domain. The classification and general mathematical description of SCNSF provide a solid theoretical foundation for further research. The proposal of three types of classic systems within SCNSF and the investigation of their chaotic properties and hardware implementation open up new avenues for practical applications. The large chaotic range exhibited by these systems implies their potential applications in various fields such as secure communication and chaotic circuit design. Moreover, the discovery of the chaos generation mechanism based on the coupling effect of the Sine-Cosine function deepens our understanding of the origin of chaos. In the complex number domain, the high parameter sensitivity and rich fractal patterns of SCNSF can be can be harnessed to develop more advanced encryption algorithms and more sensitive signal detection methods, thereby contributing to the advancement of information security and signal processing technologies. Overall, the chaotic and fractal properties of SCNSF make it a valuable asset in the pursuit of innovative solutions in multiple scientific and engineering disciplines.

Recently, deep learning methods have gained remarkable achievements in the field of image restoration for remote sensing (RS). However, most existing RS image restoration methods focus mainly on conventional first-order degradation models, which may not effectively capture the imaging mechanisms of remote sensing images. Furthermore, many RS image restoration approaches that use deep learning are often criticized for their lacks of architecture transparency and model interpretability. To address these problems, we propose a novel progressive restoration network for high-order degradation imaging (HDI-PRNet), to progressively restore different image degradation. HDI-PRNet is developed based on the theoretical framework of degradation imaging, offering the benefit of mathematical interpretability within the unfolding network. The framework is composed of three main components: a module for image denoising that relies on proximal mapping prior learning, a module for image deblurring that integrates Neumann series expansion with dual-domain degradation learning, and a module for super-resolution. Extensive experiments demonstrate that our method achieves superior performance on both synthetic and real remote sensing images., Comment: 14 pages

Face-swapping techniques have advanced rapidly with the evolution of deep learning, leading to widespread use and growing concerns about potential misuse, especially in cases of fraud. While many efforts have focused on detecting swapped face images or videos, these methods are insufficient for tracing the malicious users behind fraudulent activities. Intrusive watermark-based approaches also fail to trace unmarked identities, limiting their practical utility. To address these challenges, we introduce FaceTracer, the first non-intrusive framework specifically designed to trace the identity of the source person from swapped face images or videos. Specifically, FaceTracer leverages a disentanglement module that effectively suppresses identity information related to the target person while isolating the identity features of the source person. This allows us to extract robust identity information that can directly link the swapped face back to the original individual, aiding in uncovering the actors behind fraudulent activities. Extensive experiments demonstrate FaceTracer's effectiveness across various face-swapping techniques, successfully identifying the source person in swapped content and enabling the tracing of malicious actors involved in fraudulent activities. Additionally, FaceTracer shows strong transferability to unseen face-swapping methods including commercial applications and robustness against transmission distortions and adaptive attacks., Comment: 17 pages, 18 figures, under review

Seismic tomography is a methodology to image subsurface properties of the Earth. In order to better interpret the resulting images, it is important to assess uncertainty in the results. Mixture density networks (MDNs) provide an efficient way to estimate Bayesian posterior probability density functions (pdfs) that describe the uncertainty of tomographic images. However, the method can only be applied in cases where the number of data is fixed, and consequently a large number of practical applications that have variable data sizes cannot be solved. To resolve this issue, we introduce graph neural networks (GNNs) to solve seismic tomographic problems. Graphs are data structure which provides flexible representation of complex, variable systems. GNNs are neural networks that manipulates graph data, and can be combined with MDNs (called graph MDNs) to provide efficient estimates of posterior pdfs for graph data. In this study we apply graph MDNs to seismic tomography by representing travel time data with a graph. We demonstrate the method using both synthetic and real data, and compare the results with those obtained using Markov chain Monte Carlo (McMC). The results show that graph MDNs can provide comparable posterior pdfs to those obtained using McMC at significantly lower cost. We thus conclude that graph MDNs can be used in a range of practical applications that require many similar seismic tomographic problems with different number of data to be solved.

The studies of transuranium nuclei are of vital significance in exploring the existence of the ``island of superheavy nuclei". This work presents the systematic investigations for the ground-state properties and structure evolutions of odd-$A$ transuranium Bk isotopes taking the deformed relativistic Hartree-Bogoliubov theory in continuum (DRHBc) with PC-PK1 density functional, in comparison with those by spherical relativistic continuum Hartree-Bogoliubov (RCHB) theory. The DRHBc calculations offer improved descriptions of the binding energies, closely aligning with the experimental data. The incorporation of deformation effects in DRHBc results in enhanced nuclear binding energies and a notable reduction in $\alpha$-decay energies. With the rotational corrections further incorporated, the theoretical deviation by DRHBc from the experimental data is further reduced. Based on the two-neutron gap $\delta_{\rm 2n}$ and the neutron pairing energy $E_{\rm pair}^n$, prominent shell closures at $N=184$ and $258$, as well as potential sub-shell structures at $N=142, 150, 162, 178, 218$, and $230$ are exhibited. A quasi-periodic variation among prolate, oblate, and spherical shapes as well as prolate deformation predominance have been shown in the evolutions of the quadrupole deformation. Possible shape coexistence is predicted in $^{331}$Bk with the oblate and prolate minima in close energies, which is further supported by the triaxial relativistic-mean-field~(RMF)+BCS calculations., Comment: 9 pages, 7 figures

Personalization stands as the cornerstone of recommender systems (RecSys), striving to sift out redundant information and offer tailor-made services for users. However, the conventional cloud-based RecSys necessitates centralized data collection, posing significant risks of user privacy breaches. In response to this challenge, federated recommender systems (FedRecSys) have emerged, garnering considerable attention. FedRecSys enable users to retain personal data locally and solely share model parameters with low privacy sensitivity for global model training, significantly bolstering the system's privacy protection capabilities. Within the distributed learning framework, the pronounced non-iid nature of user behavior data introduces fresh hurdles to federated optimization. Meanwhile, the ability of federated learning to concurrently learn multiple models presents an opportunity for personalized user modeling. Consequently, the development of personalized FedRecSys (PFedRecSys) is crucial and holds substantial significance. This tutorial seeks to provide an introduction to PFedRecSys, encompassing (1) an overview of existing studies on PFedRecSys, (2) a comprehensive taxonomy of PFedRecSys spanning four pivotal research directions-client-side adaptation, server-side aggregation, communication efficiency, privacy and protection, and (3) exploration of open challenges and promising future directions in PFedRecSys. This tutorial aims to establish a robust foundation and spark new perspectives for subsequent exploration and practical implementations in the evolving realm of RecSys., Comment: A technical tutorial will appear at The Web Conference 2025

A comprehensive study of the low-temperature properties of YbNi$_4$Mg has revealed evidence of a superheavy-fermion state, characterized by a large electronic specific-heat coefficient $\gamma_0$ $\approx$ 5.65 J mol$^{-1}$ K$^{-2}$ and an elevated Wilson ratio $R_W$ = 32.1. No magnetic ordering was observed down to 70 mK; however, a broad maximum appears in the specific heat at $T^*$ = 0.3 K, along with a shoulder in the derivative of susceptibility d$\chi$/d$T$ and resistivity d$\rho$/d$T$. These features indicate a cooperative yet short-ranged magnetism entwined with the superheavy Fermi liquid. The large Wilson ratio, which is also detected in other superheavy-fermion compounds lacking long-range order, might be a signature of residual spin fluctuations. Applying a weak magnetic field of $\sim$0.1 T induces a metamagnetic-like crossover, as demonstrated by the quasi-adiabatic demagnetization measurements showing a broad minimum in the temperature-field trace. Here, an enhanced magnetocaloric cooling effect stemming from the field-sensitive superheavy-fermion state is observed, rivaling that of the well-established insulating magnetic coolants like the rare-earth garnet Gd$_3$Ga$_5$O$_{12}$.

The advent of third-generation (3G) gravitational-wave (GW) detectors opens new opportunities for multi-messenger observations of binary neutron star merger events, holding significant potential for probing the history of cosmic expansion. In this paper, we investigate the holographic dark energy (HDE) model by using the future GW standard siren data observed from the 3G GW detectors and the short $\gamma$-ray burst THESEUS-like detector joint observations. We find that GW data alone can achieve a relatively precise estimation of the Hubble constant, with precision of $0.2\%$-$0.6\%$, but its ability to constrain other cosmological parameters remains limited. Nonetheless, since the GW data can break parameter degeneracies generated by the mainstream EM observations, CMB + BAO + SN (CBS), GW standard sirens play a crucial role in enhancing the accuracy of parameter estimation. With the addition of GW data to CBS, the constraints on cosmological parameters $H_0$, $c$ and $\Omega_{\rm{m}}$ can be improved by $63\%$-$88\%$, $27\%$-$44\%$ and $55\%$-$70\%$. In summary, observations of GW standard sirens from 3G GW detectors could be pivotal in helping solve the Hubble tension and probe the fundamental nature of dark energy., Comment: 13 pages, 8 figures

This work focus on how to stabilize and lossless model compression, aiming to reduce model complexity and enhance efficiency without sacrificing performance due to compression errors. A key challenge is effectively leveraging compression errors and defining the boundaries for lossless compression to minimize model loss. i.e., compression for better. Currently, there is no systematic approach to determining this error boundary or understanding its specific impact on model performance. We propose a general \textbf{L}oss\textbf{L}ess \textbf{C}ompression theoretical framework (\textbf{LLC}), which further delineates the compression neighborhood and higher-order analysis boundaries through the total differential, thereby specifying the error range within which a model can be compressed without loss. To verify the effectiveness of LLC, we apply various compression techniques, including quantization and decomposition. Specifically, for quantization, we reformulate the classic quantization search problem as a grouped knapsack problem within the lossless neighborhood, achieving lossless quantization while improving computational efficiency. For decomposition, LLC addresses the approximation problem under low-rank constraints, automatically determining the rank for each layer and producing lossless low-rank models. We conduct extensive experiments on multiple neural network architectures on different datasets. The results show that without fancy tricks, LLC can effectively achieve lossless model compression. Our code will be made publicly., Comment: Under Review

Low-rank factorization is a popular model compression technique that minimizes the error $\delta$ between approximated and original weight matrices. Despite achieving performances close to the original models when $\delta$ is optimized, a performance discrepancy remains due to the separate optimization processes for low-rank factorization and model performance, resulting in unavoidable losses. We address this issue by introducing a novel joint optimization strategy for lossless low-rank weight factorization, which, for the first time, enhances the model's performance beyond the original. Our approach begins with a theoretical analysis of the relationship between low-rank factorization and model optimization objectives, establishing a precise perturbation range for matrix factorization errors on model performance. This challenge is then reformulated as a numerical rank deficiency problem with inequality constraints and develop a joint objective that simultaneously addresses factorization error and model performance. Based on the above analysis, we propose two optimization algorithms: \textbf{a lossless optimization algorithm} that maximizes model accuracy while ensuring compression, and \textbf{a compact optimization algorithm} that minimizes model size while preserving performance. These algorithms do not require fine-tuning and can directly compress numerous deep models to achieve lossless results. Our methods demonstrate robust efficacy across various vision and language tasks. For example, the compressed model reduced by 70\% on ResNext50 outperforms the original. Our code will be made public., Comment: Under Review

In recent years, large language models (LLMs) have been widely adopted in political science tasks such as election prediction, sentiment analysis, policy impact assessment, and misinformation detection. Meanwhile, the need to systematically understand how LLMs can further revolutionize the field also becomes urgent. In this work, we--a multidisciplinary team of researchers spanning computer science and political science--present the first principled framework termed Political-LLM to advance the comprehensive understanding of integrating LLMs into computational political science. Specifically, we first introduce a fundamental taxonomy classifying the existing explorations into two perspectives: political science and computational methodologies. In particular, from the political science perspective, we highlight the role of LLMs in automating predictive and generative tasks, simulating behavior dynamics, and improving causal inference through tools like counterfactual generation; from a computational perspective, we introduce advancements in data preparation, fine-tuning, and evaluation methods for LLMs that are tailored to political contexts. We identify key challenges and future directions, emphasizing the development of domain-specific datasets, addressing issues of bias and fairness, incorporating human expertise, and redefining evaluation criteria to align with the unique requirements of computational political science. Political-LLM seeks to serve as a guidebook for researchers to foster an informed, ethical, and impactful use of Artificial Intelligence in political science. Our online resource is available at: http://political-llm.org/., Comment: 54 Pages, 9 Figures

Post-Training Quantization (PTQ) converts pre-trained Full-Precision (FP) models into quantized versions without training. While existing methods reduce size and computational costs, they also significantly degrade performance and quantization efficiency at extremely low settings due to quantization noise. We introduce a deep model series expansion framework to address this issue, enabling rapid and accurate approximation of unquantized models without calibration sets or fine-tuning. This is the first use of series expansion for neural network quantization. Specifically, our method expands the FP model into multiple low-bit basis models. To ensure accurate quantization, we develop low-bit basis model expansions at different granularities (tensor, layer, model), and theoretically confirm their convergence to the dense model, thus restoring FP model accuracy. Additionally, we design AbelianAdd/Mul operations between isomorphic models in the low-bit expansion, forming an Abelian group to ensure operation parallelism and commutativity. The experiments show that our algorithm achieves state-of-the-art performance in low-bit settings; for example, 4-bit quantization of ResNet-50 surpasses the original accuracy, reaching 77.03%. The code will be made public., Comment: Under Review

The integration of two-dimensional (2D) van der Waals (vdW) magnets with topological insulators or heavy metals holds great potential for realizing next-generation spintronic memory devices. However, achieving high-efficiency SOT switching of monolayer vdW magnets at room temperature poses a significant challenge, particularly without an external magnetic field. Here, we show field-free, deterministic, and nonvolatile SOT switching of perpendicular magnetization in the monolayer, diluted magnetic semiconductor (DMS), Fe-doped MoS2(Fe:MoS2) at up to 380 K with a current density of $7\times10^4 A cm^{-2}$. The in situ doping of Fe into monolayer MoS2 via chemical vapor deposition and the geometry-induced strain in the crystal break the rotational switching symmetry in Fe:MoS2, promoting field-free SOT switching by generating out-of-plane spins via spin-to-spin conversion. An apparent anomalous Hall effect (AHE) loop shift at a zero in-plane magnetic field verifies the existence of z spins in Fe:MoS2, inducing an antidamping-like torque that facilitates field-free SOT switching. A strong topological Hall effect (THE) was also observed, attributed to the interfacial Dzyaloshinskii-Moriya interaction (DMI), reducing the energy barrier for SOT switching. This field-free SOT application using a 2D ferromagnetic monolayer provides a new pathway for developing highly power-efficient spintronic memory devices., Comment: 26 pages, including SOI

As language models regularly make mistakes when solving math problems, automated identification of errors in the reasoning process becomes increasingly significant for their scalable oversight. In this paper, we introduce ProcessBench for measuring the ability to identify erroneous steps in mathematical reasoning. It consists of 3,400 test cases, primarily focused on competition- and Olympiad-level math problems. Each test case contains a step-by-step solution with error location annotated by human experts. Models are required to identify the earliest step that contains an error, or conclude that all steps are correct. We conduct extensive evaluation on ProcessBench, involving two types of models: process reward models (PRMs) and critic models, where for the latter we prompt general language models to critique each solution step by step. We draw two main observations: (1) Existing PRMs typically fail to generalize to more challenging math problems beyond GSM8K and MATH. They underperform both critic models (i.e., prompted general language models) and our own trained PRM that is straightforwardly fine-tuned on the PRM800K dataset. (2) The best open-source model, QwQ-32B-Preview, has demonstrated the critique capability competitive with the proprietary model GPT-4o, despite that it still lags behind the reasoning-specialized o1-mini. We hope ProcessBench can foster future research in reasoning process assessment, paving the way toward scalable oversight of language models.

This groundbreaking research extracted DNA from petroleum using nanoparticle affinity bead technology, yielding 3,159,020 petroleum DNA (pDNA) sequences, primarily environmental DNA. While most original in situ DNA (oriDNA) was lost, ancient DNA (aDNA) from petroleum offers an important source of ecological and evolutionary information, surpassing traditional fossils. This study reveals that oil, mainly sourced from algae and lower aquatic plants, now serves as a new type of fossil, providing detailed insights into Earth's hidden history, including unclassified species and ancient events, revolutionizing petroleum geology and paleontology., Comment: 85 pages, 7 Figures, 13 Table

High quality ancient DNA (aDNA) is essential for molecular paleontology. Due to DNA degradation and contamination by environmental DNA (eDNA), current research is limited to fossils less than 1 million years old. The study successfully extracted DNA from Lycoptera davidi fossils from the Early Cretaceous period, dating 120 million years ago. Using high-throughput sequencing, 1,258,901 DNA sequences were obtained. We established a rigorous protocol known as the mega screen method. Using this method, we identified 243 original in situ DNA (oriDNA) sequences, likely from the Lycoptera genome. These sequences have an average length of over 100 base pairs and show no signs of deamination. Additionally, 10 transposase coding sequences were discovered, shedding light on a unique self-renewal mechanism in the genome. This study provides valuable DNA data for understanding ancient fish evolution and advances paleontological research., Comment: 14 pages,3 Figures

Large Language Models (LLMs) have emerged as a transformative AI paradigm, profoundly influencing daily life through their exceptional language understanding and contextual generation capabilities. Despite their remarkable performance, LLMs face a critical challenge: the propensity to produce unreliable outputs due to the inherent limitations of their learning-based nature. Formal methods (FMs), on the other hand, are a well-established computation paradigm that provides mathematically rigorous techniques for modeling, specifying, and verifying the correctness of systems. FMs have been extensively applied in mission-critical software engineering, embedded systems, and cybersecurity. However, the primary challenge impeding the deployment of FMs in real-world settings lies in their steep learning curves, the absence of user-friendly interfaces, and issues with efficiency and adaptability. This position paper outlines a roadmap for advancing the next generation of trustworthy AI systems by leveraging the mutual enhancement of LLMs and FMs. First, we illustrate how FMs, including reasoning and certification techniques, can help LLMs generate more reliable and formally certified outputs. Subsequently, we highlight how the advanced learning capabilities and adaptability of LLMs can significantly enhance the usability, efficiency, and scalability of existing FM tools. Finally, we show that unifying these two computation paradigms -- integrating the flexibility and intelligence of LLMs with the rigorous reasoning abilities of FMs -- has transformative potential for the development of trustworthy AI software systems. We acknowledge that this integration has the potential to enhance both the trustworthiness and efficiency of software engineering practices while fostering the development of intelligent FM tools capable of addressing complex yet real-world challenges., Comment: 24 pages, 4 figures

Recently, Large language models (LLMs) have revolutionized Natural Language Processing (NLP). Pretrained LLMs, due to limited training context size, struggle with handling long token sequences, limiting their performance on various downstream tasks. Current solutions toward long context modeling often employ multi-stage continual pertaining, which progressively increases the effective context length through several continual pretraining stages. However, those approaches require extensive manual tuning and human expertise. In this paper, we introduce a novel single-stage continual pretraining method, Head-Adaptive Rotary Position Encoding (HARPE), to equip LLMs with long context modeling capabilities while simplifying the training process. Our HARPE leverages different Rotary Position Encoding (RoPE) base frequency values across different attention heads and directly trains LLMs on the target context length. Extensive experiments on 4 language modeling benchmarks, including the latest RULER benchmark, demonstrate that HARPE excels in understanding and integrating long-context tasks with single-stage training, matching and even outperforming existing multi-stage methods. Our results highlight that HARPE successfully breaks the stage barrier for training LLMs with long context modeling capabilities.

Using $(27.12\pm 0.14)\times 10^{8}$ $\psi(3686)$ events collected with the BESIII detector operating at the BEPCII collider, the decay $\psi(3686)\to\Sigma^{0}\bar{\Sigma}^{0}\phi$ is observed for the first time with a statistical significance of 7.6$\sigma$. Its branching fraction is measured to be $(2.64 \pm 0.32_{\textrm{stat}} \pm 0.12_{\textrm{sys}}) \times 10^{-6}$, where the first uncertainty is statistical and the second is systematic. In addition, we search for potential intermediate states in the $\Sigma^{0}\phi$($\bar{\Sigma}^{0}\phi$) invariant mass distribution and a possible threshold enhancement in the $\Sigma^{0}\bar{\Sigma}^{0}$ system, but no conclusive evidence of is observed.