id
stringlengths
9
16
title
stringlengths
4
278
categories
stringlengths
5
104
abstract
stringlengths
6
4.09k
2502.07424
RomanLens: The Role Of Latent Romanization In Multilinguality In LLMs
cs.CL cs.AI
Large Language Models (LLMs) exhibit remarkable multilingual generalization despite being predominantly trained on English-centric corpora. A fundamental question arises: how do LLMs achieve such robust multilingual capabilities? We take the case of non-Roman script languages, we investigate the role of Romanization - the representation of non-Roman scripts using Roman characters - as a bridge in multilingual processing. Using mechanistic interpretability techniques, we analyze next-token generation and find that intermediate layers frequently represent target words in Romanized form before transitioning to native script, a phenomenon we term Latent Romanization. Further, through activation patching experiments, we demonstrate that LLMs encode semantic concepts similarly across native and Romanized scripts, suggesting a shared underlying representation. Additionally, for translation into non-Roman script languages, our findings reveal that when the target language is in Romanized form, its representations emerge earlier in the model's layers compared to native script. These insights contribute to a deeper understanding of multilingual representation in LLMs and highlight the implicit role of Romanization in facilitating language transfer.
2502.07425
Towards a Foundation Model for Physics-Informed Neural Networks: Multi-PDE Learning with Active Sampling
cs.LG
Physics-Informed Neural Networks (PINNs) have emerged as a powerful framework for solving partial differential equations (PDEs) by embedding physical laws into neural network training. However, traditional PINN models are typically designed for single PDEs, limiting their generalizability across different physical systems. In this work, we explore the potential of a foundation PINN model capable of solving multiple PDEs within a unified architecture. We investigate the efficacy of a single PINN framework trained on four distinct PDEs-the Simple Harmonic Oscillator (SHO), the 1D Heat Equation, the 1D Wave Equation, and the 2D Laplace Equation, demonstrating its ability to learn diverse physical dynamics. To enhance sample efficiency, we incorporate Active Learning (AL) using Monte Carlo (MC) Dropout-based uncertainty estimation, selecting the most informative training samples iteratively. We evaluate different active learning strategies, comparing models trained on 10%, 20%, 30%, 40%, and 50% of the full dataset, and analyze their impact on solution accuracy. Our results indicate that targeted uncertainty sampling significantly improves performance with fewer training samples, leading to efficient learning across multiple PDEs. This work highlights the feasibility of a generalizable PINN-based foundation model, capable of adapting to different physics-based problems without redesigning network architectures. Our findings suggest that multi-PDE PINNs with active learning can serve as an effective approach for reducing computational costs while maintaining high accuracy in physics-based deep learning applications.
2502.07431
ArthroPhase: A Novel Dataset and Method for Phase Recognition in Arthroscopic Video
cs.CV
This study aims to advance surgical phase recognition in arthroscopic procedures, specifically Anterior Cruciate Ligament (ACL) reconstruction, by introducing the first arthroscopy dataset and developing a novel transformer-based model. We aim to establish a benchmark for arthroscopic surgical phase recognition by leveraging spatio-temporal features to address the specific challenges of arthroscopic videos including limited field of view, occlusions, and visual distortions. We developed the ACL27 dataset, comprising 27 videos of ACL surgeries, each labeled with surgical phases. Our model employs a transformer-based architecture, utilizing temporal-aware frame-wise feature extraction through a ResNet-50 and transformer layers. This approach integrates spatio-temporal features and introduces a Surgical Progress Index (SPI) to quantify surgery progression. The model's performance was evaluated using accuracy, precision, recall, and Jaccard Index on the ACL27 and Cholec80 datasets. The proposed model achieved an overall accuracy of 72.91% on the ACL27 dataset. On the Cholec80 dataset, the model achieved a comparable performance with the state-of-the-art methods with an accuracy of 92.4%. The SPI demonstrated an output error of 10.6% and 9.86% on ACL27 and Cholec80 datasets respectively, indicating reliable surgery progression estimation. This study introduces a significant advancement in surgical phase recognition for arthroscopy, providing a comprehensive dataset and a robust transformer-based model. The results validate the model's effectiveness and generalizability, highlighting its potential to improve surgical training, real-time assistance, and operational efficiency in orthopedic surgery. The publicly available dataset and code will facilitate future research and development in this critical field.
2502.07432
CapyMOA: Efficient Machine Learning for Data Streams in Python
cs.LG
CapyMOA is an open-source library designed for efficient machine learning on streaming data. It provides a structured framework for real-time learning and evaluation, featuring a flexible data representation. CapyMOA includes an extensible architecture that allows integration with external frameworks such as MOA and PyTorch, facilitating hybrid learning approaches that combine traditional online algorithms with deep learning techniques. By emphasizing adaptability, scalability, and usability, CapyMOA allows researchers and practitioners to tackle dynamic learning challenges across various domains.
2502.07436
Optimizing Knowledge Distillation in Transformers: Enabling Multi-Head Attention without Alignment Barriers
cs.CV
Knowledge distillation (KD) in transformers often faces challenges due to misalignment in the number of attention heads between teacher and student models. Existing methods either require identical head counts or introduce projectors to bridge dimensional gaps, limiting flexibility and efficiency. We propose Squeezing-Heads Distillation (SHD), a novel approach that enables seamless knowledge transfer between models with varying head counts by compressing multi-head attention maps via efficient linear approximation. Unlike prior work, SHD eliminates alignment barriers without additional parameters or architectural modifications. Our method dynamically approximates the combined effect of multiple teacher heads into fewer student heads, preserving fine-grained attention patterns while reducing redundancy. Experiments across language (LLaMA, GPT) and vision (DiT, MDT) generative and vision (DeiT) discriminative tasks demonstrate SHD's effectiveness: it outperforms logit-based and feature-alignment KD baselines, achieving state-of-the-art results in image classification, image generation language fine-tuning, and language pre-training. The key innovations of flexible head compression, projector-free design, and linear-time complexity make SHD a versatile and scalable solution for distilling modern transformers. This work bridges a critical gap in KD, enabling efficient deployment of compact models without compromising performance.
2502.07441
SensPS: Sensing Personal Space Comfortable Distance between Human-Human Using Multimodal Sensors
cs.HC cs.AI
Personal space, also known as peripersonal space, is crucial in human social interaction, influencing comfort, communication, and social stress. Estimating and respecting personal space is essential for enhancing human-computer interaction (HCI) and smart environments. Personal space preferences vary due to individual traits, cultural background, and contextual factors. Advanced multimodal sensing technologies, including eye-tracking and wristband sensors, offer opportunities to develop adaptive systems that dynamically adjust to user comfort levels. Integrating physiological and behavioral data enables a deeper understanding of spatial interactions. This study develops a sensor-based model to estimate comfortable personal space and identifies key features influencing spatial preferences. Our findings show that multimodal sensors, particularly eye-tracking and physiological wristband data, can effectively predict personal space preferences, with eye-tracking data playing a more significant role. An experimental study involving controlled human interactions demonstrates that a Transformer-based model achieves the highest predictive accuracy (F1 score: 0.87) for estimating personal space. Eye-tracking features, such as gaze point and pupil diameter, emerge as the most significant predictors, while physiological signals from wristband sensors contribute marginally. These results highlight the potential for AI-driven personalization of social space in adaptive environments, suggesting that multimodal sensing can be leveraged to develop intelligent systems that optimize spatial arrangements in workplaces, educational institutions, and public settings. Future work should explore larger datasets, real-world applications, and additional physiological markers to enhance model robustness.
2502.07442
Hierarchical Document Parsing via Large Margin Feature Matching and Heuristics
cs.CL cs.CV
We present our solution to the AAAI-25 VRD-IU challenge, achieving first place in the competition. Our approach integrates large margin loss for improved feature discrimination and employs heuristic rules to refine hierarchical relationships. By combining a deep learning-based matching strategy with greedy algorithms, we achieve a significant boost in accuracy while maintaining computational efficiency. Our method attains an accuracy of 0.98904 on the private leaderboard, demonstrating its effectiveness in document structure parsing. Source codes are publicly available at https://github.com/ffyyytt/VRUID-AAAI-DAKiet
2502.07443
Approximating Human Strategic Reasoning with LLM-Enhanced Recursive Reasoners Leveraging Multi-agent Hypergames
cs.AI cs.GT
LLM-driven multi-agent-based simulations have been gaining traction with applications in game-theoretic and social simulations. While most implementations seek to exploit or evaluate LLM-agentic reasoning, they often do so with a weak notion of agency and simplified architectures. We implement a role-based multi-agent strategic interaction framework tailored to sophisticated recursive reasoners, providing the means for systematic in-depth development and evaluation of strategic reasoning. Our game environment is governed by the umpire responsible for facilitating games, from matchmaking through move validation to environment management. Players incorporate state-of-the-art LLMs in their decision mechanism, relying on a formal hypergame-based model of hierarchical beliefs. We use one-shot, 2-player beauty contests to evaluate the recursive reasoning capabilities of the latest LLMs, providing a comparison to an established baseline model from economics and data from human experiments. Furthermore, we introduce the foundations of an alternative semantic measure of reasoning to the k-level theory. Our experiments show that artificial reasoners can outperform the baseline model in terms of both approximating human behaviour and reaching the optimal solution.
2502.07445
Forget What You Know about LLMs Evaluations -- LLMs are Like a Chameleon
cs.CL cs.AI cs.LG
Large language models (LLMs) often appear to excel on public benchmarks, but these high scores may mask an overreliance on dataset-specific surface cues rather than true language understanding. We introduce the Chameleon Benchmark Overfit Detector (C-BOD), a meta-evaluation framework that systematically distorts benchmark prompts via a parametric transformation and detects overfitting of LLMs. By rephrasing inputs while preserving their semantic content and labels, C-BOD exposes whether a model's performance is driven by memorized patterns. Evaluated on the MMLU benchmark using 26 leading LLMs, our method reveals an average performance degradation of 2.15% under modest perturbations, with 20 out of 26 models exhibiting statistically significant differences. Notably, models with higher baseline accuracy exhibit larger performance differences under perturbation, and larger LLMs tend to be more sensitive to rephrasings indicating that both cases may overrely on fixed prompt patterns. In contrast, the Llama family and models with lower baseline accuracy show insignificant degradation, suggesting reduced dependency on superficial cues. Moreover, C-BOD's dataset- and model-agnostic design allows easy integration into training pipelines to promote more robust language understanding. Our findings challenge the community to look beyond leaderboard scores and prioritize resilience and generalization in LLM evaluation.
2502.07452
Eliciting Rational Initial Weights in Gradual Argumentation
cs.AI
Many semantics for weighted argumentation frameworks assume that each argument is associated with an initial weight. However, eliciting these initial weights poses challenges: (1) accurately providing a specific numerical value is often difficult, and (2) individuals frequently confuse initial weights with acceptability degrees in the presence of other arguments. To address these issues, we propose an elicitation pipeline that allows one to specify acceptability degree intervals for each argument. By employing gradual semantics, we can refine these intervals when they are rational, restore rationality when they are not, and ultimately identify possible initial weights for each argument.
2502.07455
RusCode: Russian Cultural Code Benchmark for Text-to-Image Generation
cs.CV cs.AI cs.CL
Text-to-image generation models have gained popularity among users around the world. However, many of these models exhibit a strong bias toward English-speaking cultures, ignoring or misrepresenting the unique characteristics of other language groups, countries, and nationalities. The lack of cultural awareness can reduce the generation quality and lead to undesirable consequences such as unintentional insult, and the spread of prejudice. In contrast to the field of natural language processing, cultural awareness in computer vision has not been explored as extensively. In this paper, we strive to reduce this gap. We propose a RusCode benchmark for evaluating the quality of text-to-image generation containing elements of the Russian cultural code. To do this, we form a list of 19 categories that best represent the features of Russian visual culture. Our final dataset consists of 1250 text prompts in Russian and their translations into English. The prompts cover a wide range of topics, including complex concepts from art, popular culture, folk traditions, famous people's names, natural objects, scientific achievements, etc. We present the results of a human evaluation of the side-by-side comparison of Russian visual concepts representations using popular generative models.
2502.07456
FedAPA: Server-side Gradient-Based Adaptive Personalized Aggregation for Federated Learning on Heterogeneous Data
cs.LG cs.CV
Personalized federated learning (PFL) tailors models to clients' unique data distributions while preserving privacy. However, existing aggregation-weight-based PFL methods often struggle with heterogeneous data, facing challenges in accuracy, computational efficiency, and communication overhead. We propose FedAPA, a novel PFL method featuring a server-side, gradient-based adaptive aggregation strategy to generate personalized models, by updating aggregation weights based on gradients of client-parameter changes with respect to the aggregation weights in a centralized manner. FedAPA guarantees theoretical convergence and achieves superior accuracy and computational efficiency compared to 10 PFL competitors across three datasets, with competitive communication overhead.
2502.07457
Bidirectional Uncertainty-Aware Region Learning for Semi-Supervised Medical Image Segmentation
cs.CV
In semi-supervised medical image segmentation, the poor quality of unlabeled data and the uncertainty in the model's predictions lead to models that inevitably produce erroneous pseudo-labels. These errors accumulate throughout model training, thereby weakening the model's performance. We found that these erroneous pseudo-labels are typically concentrated in high-uncertainty regions. Traditional methods improve performance by directly discarding pseudo-labels in these regions, but this can also result in neglecting potentially valuable training data. To alleviate this problem, we propose a bidirectional uncertainty-aware region learning strategy. In training labeled data, we focus on high-uncertainty regions, using precise label information to guide the model's learning in potentially uncontrollable areas. Meanwhile, in the training of unlabeled data, we concentrate on low-uncertainty regions to reduce the interference of erroneous pseudo-labels on the model. Through this bidirectional learning strategy, the model's overall performance has significantly improved. Extensive experiments show that our proposed method achieves significant performance improvement on different medical image segmentation tasks.
2502.07459
PerCul: A Story-Driven Cultural Evaluation of LLMs in Persian
cs.CL cs.AI cs.CY
Large language models predominantly reflect Western cultures, largely due to the dominance of English-centric training data. This imbalance presents a significant challenge, as LLMs are increasingly used across diverse contexts without adequate evaluation of their cultural competence in non-English languages, including Persian. To address this gap, we introduce PerCul, a carefully constructed dataset designed to assess the sensitivity of LLMs toward Persian culture. PerCul features story-based, multiple-choice questions that capture culturally nuanced scenarios. Unlike existing benchmarks, PerCul is curated with input from native Persian annotators to ensure authenticity and to prevent the use of translation as a shortcut. We evaluate several state-of-the-art multilingual and Persian-specific LLMs, establishing a foundation for future research in cross-cultural NLP evaluation. Our experiments demonstrate a 11.3% gap between best closed source model and layperson baseline while the gap increases to 21.3% by using the best open-weight model. You can access the dataset from here: https://huggingface.co/datasets/teias-ai/percul
2502.07460
Logarithmic Regret for Online KL-Regularized Reinforcement Learning
cs.LG stat.ML
Recent advances in Reinforcement Learning from Human Feedback (RLHF) have shown that KL-regularization plays a pivotal role in improving the efficiency of RL fine-tuning for large language models (LLMs). Despite its empirical advantage, the theoretical difference between KL-regularized RL and standard RL remains largely under-explored. While there is a recent line of work on the theoretical analysis of KL-regularized objective in decision making \citep{xiong2024iterative, xie2024exploratory,zhao2024sharp}, these analyses either reduce to the traditional RL setting or rely on strong coverage assumptions. In this paper, we propose an optimism-based KL-regularized online contextual bandit algorithm, and provide a novel analysis of its regret. By carefully leveraging the benign optimization landscape induced by the KL-regularization and the optimistic reward estimation, our algorithm achieves an $\mathcal{O}\big(\eta\log (N_{\mathcal R} T)\cdot d_{\mathcal R}\big)$ logarithmic regret bound, where $\eta, N_{\mathcal R},T,d_{\mathcal R}$ denote the KL-regularization parameter, the cardinality of the reward function class, number of rounds, and the complexity of the reward function class. Furthermore, we extend our algorithm and analysis to reinforcement learning by developing a novel decomposition over transition steps and also obtain a similar logarithmic regret bound.
2502.07461
JamendoMaxCaps: A Large Scale Music-caption Dataset with Imputed Metadata
cs.SD cs.AI
We introduce JamendoMaxCaps, a large-scale music-caption dataset featuring over 200,000 freely licensed instrumental tracks from the renowned Jamendo platform. The dataset includes captions generated by a state-of-the-art captioning model, enhanced with imputed metadata. We also introduce a retrieval system that leverages both musical features and metadata to identify similar songs, which are then used to fill in missing metadata using a local large language model (LLLM). This approach allows us to provide a more comprehensive and informative dataset for researchers working on music-language understanding tasks. We validate this approach quantitatively with five different measurements. By making the JamendoMaxCaps dataset publicly available, we provide a high-quality resource to advance research in music-language understanding tasks such as music retrieval, multimodal representation learning, and generative music models.
2502.07465
Crime Forecasting: A Spatio-temporal Analysis with Deep Learning Models
cs.LG cs.AI
This study uses deep-learning models to predict city partition crime counts on specific days. It helps police enhance surveillance, gather intelligence, and proactively prevent crimes. We formulate crime count prediction as a spatiotemporal sequence challenge, where both input data and prediction targets are spatiotemporal sequences. In order to improve the accuracy of crime forecasting, we introduce a new model that combines Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. We conducted a comparative analysis to access the effects of various data sequences, including raw and binned data, on the prediction errors of four deep learning forecasting models. Directly inputting raw crime data into the forecasting model causes high prediction errors, making the model unsuitable for real - world use. The findings indicate that the proposed CNN-LSTM model achieves optimal performance when crime data is categorized into 10 or 5 groups. Data binning can enhance forecasting model performance, but poorly defined intervals may reduce map granularity. Compared to dividing into 5 bins, binning into 10 intervals strikes an optimal balance, preserving data characteristics and surpassing raw data in predictive modelling efficacy.
2502.07466
Less is More: Masking Elements in Image Condition Features Avoids Content Leakages in Style Transfer Diffusion Models
cs.CV
Given a style-reference image as the additional image condition, text-to-image diffusion models have demonstrated impressive capabilities in generating images that possess the content of text prompts while adopting the visual style of the reference image. However, current state-of-the-art methods often struggle to disentangle content and style from style-reference images, leading to issues such as content leakages. To address this issue, we propose a masking-based method that efficiently decouples content from style without the need of tuning any model parameters. By simply masking specific elements in the style reference's image features, we uncover a critical yet under-explored principle: guiding with appropriately-selected fewer conditions (e.g., dropping several image feature elements) can efficiently avoid unwanted content flowing into the diffusion models, enhancing the style transfer performances of text-to-image diffusion models. In this paper, we validate this finding both theoretically and experimentally. Extensive experiments across various styles demonstrate the effectiveness of our masking-based method and support our theoretical results.
2502.07467
Integrated Sensing, Communication, and Over-The-Air Control of UAV Swarm Dynamics
eess.SP cs.SY eess.SY
Coordinated controlling a large UAV swarm requires significant spectrum resources due to the need for bandwidth allocation per UAV, posing a challenge in resource-limited environments. Over-the-air (OTA) control has emerged as a spectrum-efficient approach, leveraging electromagnetic superposition to form control signals at a base station (BS). However, existing OTA controllers lack sufficient optimization variables to meet UAV swarm control objectives and fail to integrate control with other BS functions like sensing. This work proposes an integrated sensing and OTA control framework (ISAC-OTA) for UAV swarm. The BS performs OTA signal construction (uplink) and dispatch (downlink) while simultaneously sensing objects. Two uplink post-processing methods are developed: a control-centric approach generating closed-form control signals via a feedback-looped OTA control problem, and a sensing-centric method mitigating transmission-induced interference for accurate object sensing. For the downlink, a non-convex problem is formulated and solved to minimize control signal dispatch (transmission) error while maintaining a minimum sensing signal-to-noise ratio (SNR). Simulation results show that the proposed ISAC-OTA controller achieves control performance comparable to the benchmark optimal control algorithm while maintaining high sensing accuracy, despite OTA transmission interference. Moreover, it eliminates the need for per-UAV bandwidth allocation, showcasing a spectrum-efficient method for cooperative control in future wireless systems.
2502.07469
5D Neural Surrogates for Nonlinear Gyrokinetic Simulations of Plasma Turbulence
physics.plasm-ph cs.AI cs.LG stat.ML
Nuclear fusion plays a pivotal role in the quest for reliable and sustainable energy production. A major roadblock to achieving commercially viable fusion power is understanding plasma turbulence, which can significantly degrade plasma confinement. Modelling turbulence is crucial to design performing plasma scenarios for next-generation reactor-class devices and current experimental machines. The nonlinear gyrokinetic equation underpinning turbulence modelling evolves a 5D distribution function over time. Solving this equation numerically is extremely expensive, requiring up to weeks for a single run to converge, making it unfeasible for iterative optimisation and control studies. In this work, we propose a method for training neural surrogates for 5D gyrokinetic simulations. Our method extends a hierarchical vision transformer to five dimensions and is trained on the 5D distribution function for the adiabatic electron approximation. We demonstrate that our model can accurately infer downstream physical quantities such as heat flux time trace and electrostatic potentials for single-step predictions two orders of magnitude faster than numerical codes. Our work paves the way towards neural surrogates for plasma turbulence simulations to accelerate deployment of commercial energy production via nuclear fusion.
2502.07470
On Event-Triggered Resilient Consensus Using Auxiliary Layer
eess.SY cs.SY
Due to its design simplicity, auxiliary layer-based resilient control is widely discussed in the literature to mitigate the effects of False Data Injection (FDI) attacks. However, the increased communication burden due to additional communication links for connecting an extra layer is often overlooked in the literature. This paper bridges this gap by considering an event-triggered approach for inter-layer communication between the physical layer (containing actual agents) and the auxiliary layer (containing virtual agents) for the resilient state consensus in a multi-agent system. We provide state-based and dynamic event-triggering mechanisms, the former being the motivation for the latter. The exclusion of Zeno behavior is established by proving positive minimum inter-event time (MIET). Extensive simulation and experimental results are provided to illustrate the proposed methodology.
2502.07472
Robotic In-Hand Manipulation for Large-Range Precise Object Movement: The RGMC Champion Solution
cs.RO
In-hand manipulation using multiple dexterous fingers is a critical robotic skill that can reduce the reliance on large arm motions, thereby saving space and energy. This letter focuses on in-grasp object movement, which refers to manipulating an object to a desired pose through only finger motions within a stable grasp. The key challenge lies in simultaneously achieving high precision and large-range movements while maintaining a constant stable grasp. To address this problem, we propose a simple and practical approach based on kinematic trajectory optimization with no need for pretraining or object geometries, which can be easily applied to novel objects in real-world scenarios. Adopting this approach, we won the championship for the in-hand manipulation track at the 9th Robotic Grasping and Manipulation Competition (RGMC) held at ICRA 2024. Implementation details, discussion, and further quantitative experimental results are presented in this letter, which aims to comprehensively evaluate our approach and share our key takeaways from the competition. Supplementary materials including video and code are available at https://rgmc-xl-team.github.io/ingrasp_manipulation .
2502.07474
ETimeline: An Extensive Timeline Generation Dataset based on Large Language Model
cs.IR
Timeline generation is of great significance for a comprehensive understanding of the development of events over time. Its goal is to organize news chronologically, which helps to identify patterns and trends that may be obscured when viewing news in isolation, making it easier to track the development of stories and understand the interrelationships between key events. Timelines are now common in various commercial products, but academic research in this area is notably scarce. Additionally, the current datasets are in need of refinement for enhanced utility and expanded coverage. In this paper, we propose ETimeline, which encompasses over $13,000$ news articles, spanning $600$ bilingual timelines across $28$ news domains. Specifically, we gather a candidate pool of more than $120,000$ news articles and employ the large language model (LLM) Pipeline to improve performance, ultimately yielding the ETimeline. The data analysis underscores the appeal of ETimeline. Additionally, we also provide the news pool data for further research and analysis. This work contributes to the advancement of timeline generation research and supports a wide range of tasks, including topic generation and event relationships. We believe that this dataset will serve as a catalyst for innovative research and bridge the gap between academia and industry in understanding the practical application of technology services. The dataset is available at https://zenodo.org/records/11392212
2502.07479
WebChecker: A Versatile EVL Plugin for Validating HTML Pages with Bootstrap Frameworks
cs.SE cs.AI
WebChecker is a plugin for Epsilon Validation Language (EVL), designed to validate both static and dynamic HTML pages utilizing frameworks like Bootstrap. By employing configurable EVL constraints, WebChecker enforces implicit rules governing HTML and CSS frameworks. The effectiveness of the plugin is demonstrated through its application on Bootstrap, the widely adopted HTML, CSS, and JavaScript framework. WebChecker comes with a set of EVL constraints to assess Bootstrap based web pages. To substantiate our claims, I present an illustrative example featuring two solutions that effectively enforce implicit rules.
2502.07480
Overfitting Regimes of Nadaraya-Watson Interpolators
cs.LG math.ST stat.ML stat.TH
In recent years, there has been much interest in understanding the generalization behavior of interpolating predictors, which overfit on noisy training data. Whereas standard analyses are concerned with whether a method is consistent or not, recent observations have shown that even inconsistent predictors can generalize well. In this work, we revisit the classic interpolating Nadaraya-Watson (NW) estimator (also known as Shepard's method), and study its generalization capabilities through this modern viewpoint. In particular, by varying a single bandwidth-like hyperparameter, we prove the existence of multiple overfitting behaviors, ranging non-monotonically from catastrophic, through benign, to tempered. Our results highlight how even classical interpolating methods can exhibit intricate generalization behaviors. Numerical experiments complement our theory, demonstrating the same phenomena.
2502.07486
Automated Road Extraction and Centreline Fitting in LiDAR Point Clouds
cs.CV
Road information extraction from 3D point clouds is useful for urban planning and traffic management. Existing methods often rely on local features and the refraction angle of lasers from kerbs, which makes them sensitive to variable kerb designs and issues in high-density areas due to data homogeneity. We propose an approach for extracting road points and fitting centrelines using a top-down view of LiDAR based ground-collected point clouds. This prospective view reduces reliance on specific kerb design and results in better road extraction. We first perform statistical outlier removal and density-based clustering to reduce noise from 3D point cloud data. Next, we perform ground point filtering using a grid-based segmentation method that adapts to diverse road scenarios and terrain characteristics. The filtered points are then projected onto a 2D plane, and the road is extracted by a skeletonisation algorithm. The skeleton is back-projected onto the 3D point cloud with calculated normals, which guide a region growing algorithm to find nearby road points. The extracted road points are then smoothed with the Savitzky-Golay filter to produce the final centreline. Our initial approach without post-processing of road skeleton achieved 67% in IoU by testing on the Perth CBD dataset with different road types. Incorporating the post-processing of the road skeleton improved the extraction of road points around the smoothed skeleton. The refined approach achieved a higher IoU value of 73% and with 23% reduction in the processing time. Our approach offers a generalised and computationally efficient solution that combines 3D and 2D processing techniques, laying the groundwork for future road reconstruction and 3D-to-2D point cloud alignment.
2502.07487
Multi-Agent Collaboration for Multilingual Code Instruction Tuning
cs.CL
Recent advancement in code understanding and generation demonstrates that code LLMs fine-tuned on a high-quality instruction dataset can gain powerful capabilities to address wide-ranging code-related tasks. However, most previous existing methods mainly view each programming language in isolation and ignore the knowledge transfer among different programming languages. To bridge the gap among different programming languages, we introduce a novel multi-agent collaboration framework to enhance multilingual instruction tuning for code LLMs, where multiple language-specific intelligent agent components with generation memory work together to transfer knowledge from one language to another efficiently and effectively. Specifically, we first generate the language-specific instruction data from the code snippets and then provide the generated data as the seed data for language-specific agents. Multiple language-specific agents discuss and collaborate to formulate a new instruction and its corresponding solution (A new programming language or existing programming language), To further encourage the cross-lingual transfer, each agent stores its generation history as memory and then summarizes its merits and faults. Finally, the high-quality multilingual instruction data is used to encourage knowledge transfer among different programming languages to train Qwen2.5-xCoder. Experimental results on multilingual programming benchmarks demonstrate the superior performance of Qwen2.5-xCoder in sharing common knowledge, highlighting its potential to reduce the cross-lingual gap.
2502.07488
Improving Adaptive Moment Optimization via Preconditioner Diagonalization
cs.LG
Modern adaptive optimization methods, such as Adam and its variants, have emerged as the most widely used tools in deep learning over recent years. These algorithms offer automatic mechanisms for dynamically adjusting the update step based on estimates of gradient statistics. Compared to traditional algorithms like Stochastic Gradient Descent, these adaptive methods are typically more robust to model scale and hyperparameter tuning. However, the gradient statistics employed by these methods often do not leverage sufficient gradient covariance information, leading to suboptimal updates in certain directions of the parameter space and potentially slower convergence. In this work, we keep track of such covariance statistics in the form of a structured preconditioner matrix. Unlike other works, our approach does not apply direct approximations to estimate this matrix. We instead implement an invertible transformation that maps the preconditioner matrix into a new space where it becomes approximately diagonal. This enables a diagonal approximation of the preconditioner matrix in the transformed space, offering several computational advantages. Empirical results show that our approach can substantially enhance the convergence speed of modern adaptive optimizers. Notably, for large language models like LLaMA, we can achieve a speedup of 2x compared to the baseline Adam. Additionally, our method can be integrated with memory-efficient optimizers like Adafactor to manage computational overhead.
2502.07489
Physiome-ODE: A Benchmark for Irregularly Sampled Multivariate Time Series Forecasting Based on Biological ODEs
cs.LG
State-of-the-art methods for forecasting irregularly sampled time series with missing values predominantly rely on just four datasets and a few small toy examples for evaluation. While ordinary differential equations (ODE) are the prevalent models in science and engineering, a baseline model that forecasts a constant value outperforms ODE-based models from the last five years on three of these existing datasets. This unintuitive finding hampers further research on ODE-based models, a more plausible model family. In this paper, we develop a methodology to generate irregularly sampled multivariate time series (IMTS) datasets from ordinary differential equations and to select challenging instances via rejection sampling. Using this methodology, we create Physiome-ODE, a large and sophisticated benchmark of IMTS datasets consisting of 50 individual datasets, derived from real-world ordinary differential equations from research in biology. Physiome-ODE is the first benchmark for IMTS forecasting that we are aware of and an order of magnitude larger than the current evaluation setting of four datasets. Using our benchmark Physiome-ODE, we show qualitatively completely different results than those derived from the current four datasets: on Physiome-ODE ODE-based models can play to their strength and our benchmark can differentiate in a meaningful way between different IMTS forecasting models. This way, we expect to give a new impulse to research on ODE-based time series modeling.
2502.07490
Mask-Enhanced Autoregressive Prediction: Pay Less Attention to Learn More
cs.CL cs.LG
Large Language Models (LLMs) are discovered to suffer from accurately retrieving key information. To address this, we propose Mask-Enhanced Autoregressive Prediction (MEAP), a simple yet effective training paradigm that seamlessly integrates Masked Language Modeling (MLM) into Next-Token Prediction (NTP) to enhance the latter's in-context retrieval capabilities. Specifically, MEAP first randomly masks a small fraction of input tokens and then directly performs the standard next-token prediction autoregressive using a decoder-only Transformer. MEAP eliminates the need for bidirectional attention or encoder-decoder architectures for MLM, incurring no additional computational overhead during pre-training or inference. Intensive experiments demonstrate that MEAP substantially outperforms NTP on key information retrieval and long-context reasoning tasks, while performing on par or better on commonsense reasoning tasks. The benefits of MEAP also extend to supervised fine-tuning, where it shows remarkable advantages in lost-in-the-middle scenarios, outperforming NTP by 11.77 percentage points. Our analysis indicates that MEAP's effectiveness arises from its ability to promote more distinguishable attention scores by concentrating on a reduced set of non-masked tokens. This mechanism improves the model's focus on task-relevant signals while mitigating the influence of peripheral context. These findings position MEAP as a promising training paradigm for large language models.
2502.07491
Exploring Patterns Behind Sports
cs.LG cs.IR
This paper presents a comprehensive framework for time series prediction using a hybrid model that combines ARIMA and LSTM. The model incorporates feature engineering techniques, including embedding and PCA, to transform raw data into a lower-dimensional representation while retaining key information. The embedding technique is used to convert categorical data into continuous vectors, facilitating the capture of complex relationships. PCA is applied to reduce dimensionality and extract principal components, enhancing model performance and computational efficiency. To handle both linear and nonlinear patterns in the data, the ARIMA model captures linear trends, while the LSTM model models complex nonlinear dependencies. The hybrid model is trained on historical data and achieves high accuracy, as demonstrated by low RMSE and MAE scores. Additionally, the paper employs the run test to assess the randomness of sequences, providing insights into the underlying patterns. Ablation studies are conducted to validate the roles of different components in the model, demonstrating the significance of each module. The paper also utilizes the SHAP method to quantify the impact of traditional advantages on the predicted results, offering a detailed understanding of feature importance. The KNN method is used to determine the optimal prediction interval, further enhancing the model's accuracy. The results highlight the effectiveness of combining traditional statistical methods with modern deep learning techniques for robust time series forecasting in Sports.
2502.07492
RoMA: Robust Malware Attribution via Byte-level Adversarial Training with Global Perturbations and Adversarial Consistency Regularization
cs.CR cs.CV
Attributing APT (Advanced Persistent Threat) malware to their respective groups is crucial for threat intelligence and cybersecurity. However, APT adversaries often conceal their identities, rendering attribution inherently adversarial. Existing machine learning-based attribution models, while effective, remain highly vulnerable to adversarial attacks. For example, the state-of-the-art byte-level model MalConv sees its accuracy drop from over 90% to below 2% under PGD (projected gradient descent) attacks. Existing gradient-based adversarial training techniques for malware detection or image processing were applied to malware attribution in this study, revealing that both robustness and training efficiency require significant improvement. To address this, we propose RoMA, a novel single-step adversarial training approach that integrates global perturbations to generate enhanced adversarial samples and employs adversarial consistency regularization to improve representation quality and resilience. A novel APT malware dataset named AMG18, with diverse samples and realistic class imbalances, is introduced for evaluation. Extensive experiments show that RoMA significantly outperforms seven competing methods in both adversarial robustness (e.g., achieving over 80% robust accuracy-more than twice that of the next-best method under PGD attacks) and training efficiency (e.g., more than twice as fast as the second-best method in terms of accuracy), while maintaining superior standard accuracy in non-adversarial scenarios.
2502.07494
URECA: The Chain of Two Minimum Set Cover Problems exists behind Adaptation to Shifts in Semantic Code Search
cs.AI
Adaptation is to make model learn the patterns shifted from the training distribution. In general, this adaptation is formulated as the minimum entropy problem. However, the minimum entropy problem has inherent limitation -- shifted initialization cascade phenomenon. We extend the relationship between the minimum entropy problem and the minimum set cover problem via Lebesgue integral. This extension reveals that internal mechanism of the minimum entropy problem ignores the relationship between disentangled representations, which leads to shifted initialization cascade. From the analysis, we introduce a new clustering algorithm, Union-find based Recursive Clustering Algorithm~(URECA). URECA is an efficient clustering algorithm for the leverage of the relationships between disentangled representations. The update rule of URECA depends on Thresholdly-Updatable Stationary Assumption to dynamics as a released version of Stationary Assumption. This assumption helps URECA to transport disentangled representations with no errors based on the relationships between disentangled representations. URECA also utilize simulation trick to efficiently cluster disentangled representations. The wide range of evaluations show that URECA achieves consistent performance gains for the few-shot adaptation to diverse types of shifts along with advancement to State-of-The-Art performance in CoSQA in the scenario of query shift.
2502.07495
LLM-Sketch: Enhancing Network Sketches with LLM
cs.NI cs.LG
Network stream mining is fundamental to many network operations. Sketches, as compact data structures that offer low memory overhead with bounded accuracy, have emerged as a promising solution for network stream mining. Recent studies attempt to optimize sketches using machine learning; however, these approaches face the challenges of lacking adaptivity to dynamic networks and incurring high training costs. In this paper, we propose LLM-Sketch, based on the insight that fields beyond the flow IDs in packet headers can also help infer flow sizes. By using a two-tier data structure and separately recording large and small flows, LLM-Sketch improves accuracy while minimizing memory usage. Furthermore, it leverages fine-tuned large language models (LLMs) to reliably estimate flow sizes. We evaluate LLM-Sketch on three representative tasks, and the results demonstrate that LLM-Sketch outperforms state-of-the-art methods by achieving a $7.5\times$ accuracy improvement.
2502.07497
On Training-Conditional Conformal Prediction and Binomial Proportion Confidence Intervals
cs.LG
Estimating the expectation of a Bernoulli random variable based on N independent trials is a classical problem in statistics, typically addressed using Binomial Proportion Confidence Intervals (BPCI). In the control systems community, many critical tasks-such as certifying the statistical safety of dynamical systems-can be formulated as BPCI problems. Conformal Prediction (CP), a distribution-free technique for uncertainty quantification, has gained significant attention in recent years and has been applied to various control systems problems, particularly to address uncertainties in learned dynamics or controllers. A variant known as training-conditional CP was recently employed to tackle the problem of safety certification. In this note, we highlight that the use of training-conditional CP in this context does not provide valid safety guarantees. We demonstrate why CP is unsuitable for BPCI problems and argue that traditional BPCI methods are better suited for statistical safety certification.
2502.07500
Unified Graph Networks (UGN): A Deep Neural Framework for Solving Graph Problems
cs.LG
Deep neural networks have enabled researchers to create powerful generalized frameworks, such as transformers, that can be used to solve well-studied problems in various application domains, such as text and image. However, such generalized frameworks are not available for solving graph problems. Graph structures are ubiquitous in many applications around us and many graph problems have been widely studied over years. In recent times, there has been a surge in deep neural network based approaches to solve graph problems, with growing availability of graph structured datasets across diverse domains. Nevertheless, existing methods are mostly tailored to solve a specific task and lack the capability to create a generalized model leading to solutions for different downstream tasks. In this work, we propose a novel, resource-efficient framework named \emph{U}nified \emph{G}raph \emph{N}etwork (UGN) by leveraging the feature extraction capability of graph convolutional neural networks (GCN) and 2-dimensional convolutional neural networks (Conv2D). UGN unifies various graph learning tasks, such as link prediction, node classification, community detection, graph-to-graph translation, knowledge graph completion, and more, within a cohesive framework, while exercising minimal task-specific extensions (e.g., formation of supernodes for coarsening massive networks to increase scalability, use of \textit{mean target connectivity matrix} (MTCM) representation for achieving scalability in graph translation task, etc.) to enhance the generalization capability of graph learning and analysis. We test the novel UGN framework for six uncorrelated graph problems, using twelve different datasets. Experimental results show that UGN outperforms the state-of-the-art baselines by a significant margin on ten datasets, while producing comparable results on the remaining dataset.
2502.07503
Recursive Inference Scaling: A Winning Path to Scalable Inference in Language and Multimodal Systems
cs.AI cs.LG
Recent research in language modeling reveals two scaling effects: the well-known improvement from increased training compute, and a lesser-known boost from applying more sophisticated or computationally intensive inference methods. Inspired by recent findings on the fractal geometry of language, we introduce Recursive INference Scaling (RINS) as a complementary, plug-in recipe for scaling inference time. For a given fixed model architecture and training compute budget, RINS substantially improves language modeling performance. It also generalizes beyond pure language tasks, delivering gains in multimodal systems, including a +2% improvement in 0-shot ImageNet accuracy for SigLIP-B/16. Additionally, by deriving data scaling laws, we show that RINS improves both the asymptotic performance limits and the scaling exponents. These advantages are maintained even when compared to state-of-the-art recursive techniques like the "repeat-all-over" (RAO) strategy in Mobile LLM. Finally, stochastic RINS not only can enhance performance further but also provides the flexibility to optionally forgo increased inference computation at test time with minimal performance degradation.
2502.07505
Efficient Continuous Group Convolutions for Local SE(3) Equivariance in 3D Point Clouds
cs.CV
Extending the translation equivariance property of convolutional neural networks to larger symmetry groups has been shown to reduce sample complexity and enable more discriminative feature learning. Further, exploiting additional symmetries facilitates greater weight sharing than standard convolutions, leading to an enhanced network expressivity without an increase in parameter count. However, extending the equivariant properties of a convolution layer comes at a computational cost. In particular, for 3D data, expanding equivariance to the SE(3) group (rotation and translation) results in a 6D convolution operation, which is not tractable for larger data samples such as 3D scene scans. While efforts have been made to develop efficient SE(3) equivariant networks, existing approaches rely on discretization or only introduce global rotation equivariance. This limits their applicability to point clouds representing a scene composed of multiple objects. This work presents an efficient, continuous, and local SE(3) equivariant convolution layer for point cloud processing based on general group convolution and local reference frames. Our experiments show that our approach achieves competitive or superior performance across a range of datasets and tasks, including object classification and semantic segmentation, with negligible computational overhead.
2502.07508
Enhance-A-Video: Better Generated Video for Free
cs.CV
DiT-based video generation has achieved remarkable results, but research into enhancing existing models remains relatively unexplored. In this work, we introduce a training-free approach to enhance the coherence and quality of DiT-based generated videos, named Enhance-A-Video. The core idea is enhancing the cross-frame correlations based on non-diagonal temporal attention distributions. Thanks to its simple design, our approach can be easily applied to most DiT-based video generation frameworks without any retraining or fine-tuning. Across various DiT-based video generation models, our approach demonstrates promising improvements in both temporal consistency and visual quality. We hope this research can inspire future explorations in video generation enhancement.
2502.07509
Dual Arm Steering of Deformable Linear Objects in 2-D and 3-D Environments Using Euler's Elastica Solutions
cs.RO cs.SY eess.SY
This paper describes a method for steering deformable linear objects using two robot hands in environments populated by sparsely spaced obstacles. The approach involves manipulating an elastic inextensible rod by varying the gripping endpoint positions and tangents. Closed form solutions that describe the flexible linear object shape in planar environments, Euler's elastica, are described. The paper uses these solutions to formulate criteria for non self-intersection, stability and obstacle avoidance. These criteria are formulated as constraints in the flexible object six-dimensional configuration space that represents the robot gripping endpoint positions and tangents. In particular, this paper introduces a novel criterion that ensures the flexible object stability during steering. All safety criteria are integrated into a scheme for steering flexible linear objects in planar environments, which is lifted into a steering scheme in three-dimensional environments populated by sparsely spaced obstacles. Experiments with a dual-arm robot demonstrate the method.
2502.07510
Joint Metric Space Embedding by Unbalanced OT with Gromov-Wasserstein Marginal Penalization
cs.LG
We propose a new approach for unsupervised alignment of heterogeneous datasets, which maps data from two different domains without any known correspondences to a common metric space. Our method is based on an unbalanced optimal transport problem with Gromov-Wasserstein marginal penalization. It can be seen as a counterpart to the recently introduced joint multidimensional scaling method. We prove that there exists a minimizer of our functional and that for penalization parameters going to infinity, the corresponding sequence of minimizers converges to a minimizer of the so-called embedded Wasserstein distance. Our model can be reformulated as a quadratic, multi-marginal, unbalanced optimal transport problem, for which a bi-convex relaxation admits a numerical solver via block-coordinate descent. We provide numerical examples for joint embeddings in Euclidean as well as non-Euclidean spaces.
2502.07511
Quantitative evaluation of unsupervised clustering algorithms for dynamic total-body PET image analysis
stat.AP cs.CV
Background. Recently, dynamic total-body positron emission tomography (PET) imaging has become possible due to new scanner devices. While clustering algorithms have been proposed for PET analysis already earlier, there is still little research systematically evaluating these algorithms for processing of dynamic total-body PET images. Materials and methods. Here, we compare the performance of 15 unsupervised clustering methods, including K-means either by itself or after principal component analysis (PCA) or independent component analysis (ICA), Gaussian mixture model (GMM), fuzzy c-means (FCM), agglomerative clustering, spectral clustering, and several newer clustering algorithms, for classifying time activity curves (TACs) in dynamic PET images. We use dynamic total-body $^{15}$O-water PET images collected from 30 patients with suspected or confirmed coronary artery disease. To evaluate the clustering algorithms in a quantitative way, we use them to classify 5000 TACs from each image based on whether the curve is taken from brain, right heart ventricle, right kidney, lower right lung lobe, or urinary bladder. Results. According to our results, the best methods are GMM, FCM, and ICA combined with mini batch K-means, which classified the TACs with a median accuracies of 89\%, 83\%, and 81\%, respectively, in a processing time of half a second or less on average for each image. Conclusion. GMM, FCM, and ICA with mini batch K-means show promise for dynamic total-body PET analysis.
2502.07514
A Near-optimal, Scalable and Corruption-tolerant Framework for Stochastic Bandits: From Single-Agent to Multi-Agent and Beyond
cs.LG
We investigate various stochastic bandit problems in the presence of adversarial corruption. A seminal contribution to this area is the BARBAR~\citep{gupta2019better} algorithm, which is both simple and efficient, tolerating significant levels of corruption with nearly no degradation in performance. However, its regret upper bound exhibits a complexity of $O(KC)$, while the lower bound is $\Omega(C)$. In this paper, we enhance the BARBAR algorithm by proposing a novel framework called BARBAT, which eliminates the factor of $K$ and achieves an optimal regret bound up to a logarithmic factor. We also demonstrate how BARBAT can be extended to various settings, including graph bandits, combinatorial semi-bandits, batched bandits and multi-agent bandits. In comparison to the Follow-The-Regularized-Leader (FTRL) family of methods, which provide a best-of-both-worlds guarantee, our approach is more efficient and parallelizable. Notably, FTRL-based methods face challenges in scaling to batched and multi-agent settings.
2502.07516
The Devil is in the Prompts: De-Identification Traces Enhance Memorization Risks in Synthetic Chest X-Ray Generation
eess.IV cs.AI cs.CV cs.LG
Generative models, particularly text-to-image (T2I) diffusion models, play a crucial role in medical image analysis. However, these models are prone to training data memorization, posing significant risks to patient privacy. Synthetic chest X-ray generation is one of the most common applications in medical image analysis with the MIMIC-CXR dataset serving as the primary data repository for this task. This study presents the first systematic attempt to identify prompts and text tokens in MIMIC-CXR that contribute the most to training data memorization. Our analysis reveals two unexpected findings: (1) prompts containing traces of de-identification procedures (markers introduced to hide Protected Health Information) are the most memorized, and (2) among all tokens, de-identification markers contribute the most towards memorization. This highlights a broader issue with the standard anonymization practices and T2I synthesis with MIMIC-CXR. To exacerbate, existing inference-time memorization mitigation strategies are ineffective and fail to sufficiently reduce the model's reliance on memorized text tokens. On this front, we propose actionable strategies for different stakeholders to enhance privacy and improve the reliability of generative models in medical imaging. Finally, our results provide a foundation for future work on developing and benchmarking memorization mitigation techniques for synthetic chest X-ray generation using the MIMIC-CXR dataset. The anonymized code is available at https://anonymous.4open.science/r/diffusion_memorization-8011/
2502.07523
Scaling Off-Policy Reinforcement Learning with Batch and Weight Normalization
cs.LG cs.AI
Reinforcement learning has achieved significant milestones, but sample efficiency remains a bottleneck for real-world applications. Recently, CrossQ has demonstrated state-of-the-art sample efficiency with a low update-to-data (UTD) ratio of 1. In this work, we explore CrossQ's scaling behavior with higher UTD ratios. We identify challenges in the training dynamics, which are emphasized by higher UTD ratios. To address these, we integrate weight normalization into the CrossQ framework, a solution that stabilizes training, has been shown to prevent potential loss of plasticity and keeps the effective learning rate constant. Our proposed approach reliably scales with increasing UTD ratios, achieving competitive performance across 25 challenging continuous control tasks on the DeepMind Control Suite and Myosuite benchmarks, notably the complex dog and humanoid environments. This work eliminates the need for drastic interventions, such as network resets, and offers a simple yet robust pathway for improving sample efficiency and scalability in model-free reinforcement learning.
2502.07526
CodePhys: Robust Video-based Remote Physiological Measurement through Latent Codebook Querying
cs.CV
Remote photoplethysmography (rPPG) aims to measure non-contact physiological signals from facial videos, which has shown great potential in many applications. Most existing methods directly extract video-based rPPG features by designing neural networks for heart rate estimation. Although they can achieve acceptable results, the recovery of rPPG signal faces intractable challenges when interference from real-world scenarios takes place on facial video. Specifically, facial videos are inevitably affected by non-physiological factors (e.g., camera device noise, defocus, and motion blur), leading to the distortion of extracted rPPG signals. Recent rPPG extraction methods are easily affected by interference and degradation, resulting in noisy rPPG signals. In this paper, we propose a novel method named CodePhys, which innovatively treats rPPG measurement as a code query task in a noise-free proxy space (i.e., codebook) constructed by ground-truth PPG signals. We consider noisy rPPG features as queries and generate high-fidelity rPPG features by matching them with noise-free PPG features from the codebook. Our approach also incorporates a spatial-aware encoder network with a spatial attention mechanism to highlight physiologically active areas and uses a distillation loss to reduce the influence of non-periodic visual interference. Experimental results on four benchmark datasets demonstrate that CodePhys outperforms state-of-the-art methods in both intra-dataset and cross-dataset settings.
2502.07527
NatureLM: Deciphering the Language of Nature for Scientific Discovery
cs.AI cs.LG
Foundation models have revolutionized natural language processing and artificial intelligence, significantly enhancing how machines comprehend and generate human languages. Inspired by the success of these foundation models, researchers have developed foundation models for individual scientific domains, including small molecules, materials, proteins, DNA, and RNA. However, these models are typically trained in isolation, lacking the ability to integrate across different scientific domains. Recognizing that entities within these domains can all be represented as sequences, which together form the "language of nature", we introduce Nature Language Model (briefly, NatureLM), a sequence-based science foundation model designed for scientific discovery. Pre-trained with data from multiple scientific domains, NatureLM offers a unified, versatile model that enables various applications including: (i) generating and optimizing small molecules, proteins, RNA, and materials using text instructions; (ii) cross-domain generation/design, such as protein-to-molecule and protein-to-RNA generation; and (iii) achieving state-of-the-art performance in tasks like SMILES-to-IUPAC translation and retrosynthesis on USPTO-50k. NatureLM offers a promising generalist approach for various scientific tasks, including drug discovery (hit generation/optimization, ADMET optimization, synthesis), novel material design, and the development of therapeutic proteins or nucleotides. We have developed NatureLM models in different sizes (1 billion, 8 billion, and 46.7 billion parameters) and observed a clear improvement in performance as the model size increases.
2502.07528
Forecasting the future development in quality and value of professional football players for applications in team management
stat.AP cs.LG
Transfers in professional football (soccer) are risky investments because of the large transfer fees and high risks involved. Although data-driven models can be used to improve transfer decisions, existing models focus on describing players' historical progress, leaving their future performance unknown. Moreover, recent developments have called for the use of explainable models combined with uncertainty quantification of predictions. This paper assesses explainable machine learning models based on predictive accuracy and uncertainty quantification methods for the prediction of the future development in quality and transfer value of professional football players. Using a historical data set of data-driven indicators describing player quality and the transfer value of a football player, the models are trained to forecast player quality and player value one year ahead. These two prediction problems demonstrate the efficacy of tree-based models, particularly random forest and XGBoost, in making accurate predictions. In general, the random forest model is found to be the most suitable model because it provides accurate predictions as well as an uncertainty quantification method that naturally arises from the bagging procedure of the random forest model. Additionally, our research shows that the development of player performance contains nonlinear patterns and interactions between variables, and that time series information can provide useful information for the modeling of player performance metrics. Our research provides models to help football clubs make more informed, data-driven transfer decisions by forecasting player quality and transfer value.
2502.07529
Training Deep Learning Models with Norm-Constrained LMOs
cs.LG math.OC
In this work, we study optimization methods that leverage the linear minimization oracle (LMO) over a norm-ball. We propose a new stochastic family of algorithms that uses the LMO to adapt to the geometry of the problem and, perhaps surprisingly, show that they can be applied to unconstrained problems. The resulting update rule unifies several existing optimization methods under a single framework. Furthermore, we propose an explicit choice of norm for deep architectures, which, as a side benefit, leads to the transferability of hyperparameters across model sizes. Experimentally, we demonstrate significant speedups on nanoGPT training without any reliance on Adam. The proposed method is memory-efficient, requiring only one set of model weights and one set of gradients, which can be stored in half-precision.
2502.07531
VidCRAFT3: Camera, Object, and Lighting Control for Image-to-Video Generation
cs.CV cs.AI cs.LG cs.MM
Recent image-to-video generation methods have demonstrated success in enabling control over one or two visual elements, such as camera trajectory or object motion. However, these methods are unable to offer control over multiple visual elements due to limitations in data and network efficacy. In this paper, we introduce VidCRAFT3, a novel framework for precise image-to-video generation that enables control over camera motion, object motion, and lighting direction simultaneously. To better decouple control over each visual element, we propose the Spatial Triple-Attention Transformer, which integrates lighting direction, text, and image in a symmetric way. Since most real-world video datasets lack lighting annotations, we construct a high-quality synthetic video dataset, the VideoLightingDirection (VLD) dataset. This dataset includes lighting direction annotations and objects of diverse appearance, enabling VidCRAFT3 to effectively handle strong light transmission and reflection effects. Additionally, we propose a three-stage training strategy that eliminates the need for training data annotated with multiple visual elements (camera motion, object motion, and lighting direction) simultaneously. Extensive experiments on benchmark datasets demonstrate the efficacy of VidCRAFT3 in producing high-quality video content, surpassing existing state-of-the-art methods in terms of control granularity and visual coherence. All code and data will be publicly available.
2502.07532
Diffusion-LAM: Probabilistic Limited Area Weather Forecasting with Diffusion
cs.LG physics.ao-ph
Machine learning methods have been shown to be effective for weather forecasting, based on the speed and accuracy compared to traditional numerical models. While early efforts primarily concentrated on deterministic predictions, the field has increasingly shifted toward probabilistic forecasting to better capture the forecast uncertainty. Most machine learning-based models have been designed for global-scale predictions, with only limited work targeting regional or limited area forecasting, which allows more specialized and flexible modeling for specific locations. This work introduces Diffusion-LAM, a probabilistic limited area weather model leveraging conditional diffusion. By conditioning on boundary data from surrounding regions, our approach generates forecasts within a defined area. Experimental results on the MEPS limited area dataset demonstrate the potential of Diffusion-LAM to deliver accurate probabilistic forecasts, highlighting its promise for limited-area weather prediction.
2502.07541
Corporate Greenwashing Detection in Text -- a Survey
cs.CL
Greenwashing is an effort to mislead the public about the environmental impact of an entity, such as a state or company. We provide a comprehensive survey of the scientific literature addressing natural language processing methods to identify potentially misleading climate-related corporate communications, indicative of greenwashing. We break the detection of greenwashing into intermediate tasks, and review the state-of-the-art approaches for each of them. We discuss datasets, methods, and results, as well as limitations and open challenges. We also provide an overview of how far the field has come as a whole, and point out future research directions.
2502.07542
Exoplanet Transit Candidate Identification in TESS Full-Frame Images via a Transformer-Based Algorithm
astro-ph.EP astro-ph.GA astro-ph.IM cs.AI
The Transiting Exoplanet Survey Satellite (TESS) is surveying a large fraction of the sky, generating a vast database of photometric time series data that requires thorough analysis to identify exoplanetary transit signals. Automated learning approaches have been successfully applied to identify transit signals. However, most existing methods focus on the classification and validation of candidates, while few efforts have explored new techniques for the search of candidates. To search for new exoplanet transit candidates, we propose an approach to identify exoplanet transit signals without the need for phase folding or assuming periodicity in the transit signals, such as those observed in multi-transit light curves. To achieve this, we implement a new neural network inspired by Transformers to directly process Full Frame Image (FFI) light curves to detect exoplanet transits. Transformers, originally developed for natural language processing, have recently demonstrated significant success in capturing long-range dependencies compared to previous approaches focused on sequential data. This ability allows us to employ multi-head self-attention to identify exoplanet transit signals directly from the complete light curves, combined with background and centroid time series, without requiring prior transit parameters. The network is trained to learn characteristics of the transit signal, like the dip shape, which helps distinguish planetary transits from other variability sources. Our model successfully identified 214 new planetary system candidates, including 122 multi-transit light curves, 88 single-transit and 4 multi-planet systems from TESS sectors 1-26 with a radius > 0.27 $R_{\mathrm{Jupiter}}$, demonstrating its ability to detect transits regardless of their periodicity.
2502.07544
Grammar Control in Dialogue Response Generation for Language Learning Chatbots
cs.CL
Chatbots based on large language models offer cheap conversation practice opportunities for language learners. However, they are hard to control for linguistic forms that correspond to learners' current needs, such as grammar. We control grammar in chatbot conversation practice by grounding a dialogue response generation model in a pedagogical repository of grammar skills. We also explore how this control helps learners to produce specific grammar. We comprehensively evaluate prompting, fine-tuning, and decoding strategies for grammar-controlled dialogue response generation. Strategically decoding Llama3 outperforms GPT-3.5 when tolerating minor response quality losses. Our simulation predicts grammar-controlled responses to support grammar acquisition adapted to learner proficiency. Existing language learning chatbots and research on second language acquisition benefit from these affordances. Code available on GitHub.
2502.07547
Instance-dependent Early Stopping
cs.LG
In machine learning practice, early stopping has been widely used to regularize models and can save computational costs by halting the training process when the model's performance on a validation set stops improving. However, conventional early stopping applies the same stopping criterion to all instances without considering their individual learning statuses, which leads to redundant computations on instances that are already well-learned. To further improve the efficiency, we propose an Instance-dependent Early Stopping (IES) method that adapts the early stopping mechanism from the entire training set to the instance level, based on the core principle that once the model has mastered an instance, the training on it should stop. IES considers an instance as mastered if the second-order differences of its loss value remain within a small range around zero. This offers a more consistent measure of an instance's learning status compared with directly using the loss value, and thus allows for a unified threshold to determine when an instance can be excluded from further backpropagation. We show that excluding mastered instances from backpropagation can increase the gradient norms, thereby accelerating the decrease of the training loss and speeding up the training process. Extensive experiments on benchmarks demonstrate that IES method can reduce backpropagation instances by 10%-50% while maintaining or even slightly improving the test accuracy and transfer learning performance of a model.
2502.07549
HGTUL: A Hypergraph-based Model For Trajectory User Linking
cs.LG cs.AI
Trajectory User Linking (TUL), which links anonymous trajectories with users who generate them, plays a crucial role in modeling human mobility. Despite significant advancements in this field, existing studies primarily neglect the high-order inter-trajectory relationships, which represent complex associations among multiple trajectories, manifested through multi-location co-occurrence patterns emerging when trajectories intersect at various Points of Interest (POIs). Furthermore, they also overlook the variable influence of POIs on different trajectories, as well as the user class imbalance problem caused by disparities in user activity levels and check-in frequencies. To address these limitations, we propose a novel HyperGraph-based multi-perspective Trajectory User Linking model (HGTUL). Our model learns trajectory representations from both relational and spatio-temporal perspectives: (1) it captures high-order associations among trajectories by constructing a trajectory hypergraph and leverages a hypergraph attention network to learn the variable impact of POIs on trajectories; (2) it models the spatio-temporal characteristics of trajectories by incorporating their temporal and spatial information into a sequential encoder. Moreover, we design a data balancing method to effectively address the user class imbalance problem and experimentally validate its significance in TUL. Extensive experiments on three real-world datasets demonstrate that HGTUL outperforms state-of-the-art baselines, achieving improvements of 2.57%~20.09% and 5.68%~26.00% in ACC@1 and Macro-F1 metrics, respectively.
2502.07551
Early Stopping Against Label Noise Without Validation Data
cs.LG
Early stopping methods in deep learning face the challenge of balancing the volume of training and validation data, especially in the presence of label noise. Concretely, sparing more data for validation from training data would limit the performance of the learned model, yet insufficient validation data could result in a sub-optimal selection of the desired model. In this paper, we propose a novel early stopping method called Label Wave, which does not require validation data for selecting the desired model in the presence of label noise. It works by tracking the changes in the model's predictions on the training set during the training process, aiming to halt training before the model unduly fits mislabeled data. This method is empirically supported by our observation that minimum fluctuations in predictions typically occur at the training epoch before the model excessively fits mislabeled data. Through extensive experiments, we show both the effectiveness of the Label Wave method across various settings and its capability to enhance the performance of existing methods for learning with noisy labels.
2502.07552
Unsupervised Translation of Emergent Communication
cs.CL cs.AI
Emergent Communication (EC) provides a unique window into the language systems that emerge autonomously when agents are trained to jointly achieve shared goals. However, it is difficult to interpret EC and evaluate its relationship with natural languages (NL). This study employs unsupervised neural machine translation (UNMT) techniques to decipher ECs formed during referential games with varying task complexities, influenced by the semantic diversity of the environment. Our findings demonstrate UNMT's potential to translate EC, illustrating that task complexity characterized by semantic diversity enhances EC translatability, while higher task complexity with constrained semantic variability exhibits pragmatic EC, which, although challenging to interpret, remains suitable for translation. This research marks the first attempt, to our knowledge, to translate EC without the aid of parallel data.
2502.07553
Attention Learning is Needed to Efficiently Learn Parity Function
cs.LG
Transformers, with their attention mechanisms, have emerged as the state-of-the-art architectures of sequential modeling and empirically outperform feed-forward neural networks (FFNNs) across many fields, such as natural language processing and computer vision. However, their generalization ability, particularly for low-sensitivity functions, remains less studied. We bridge this gap by analyzing transformers on the $k$-parity problem. Daniely and Malach (NeurIPS 2020) show that FFNNs with one hidden layer and $O(nk^7 \log k)$ parameters can learn $k$-parity, where the input length $n$ is typically much larger than $k$. In this paper, we prove that FFNNs require at least $\Omega(n)$ parameters to learn $k$-parity, while transformers require only $O(k)$ parameters, surpassing the theoretical lower bound needed by FFNNs. We further prove that this parameter efficiency cannot be achieved with fixed attention heads. Our work establishes transformers as theoretically superior to FFNNs in learning parity function, showing how their attention mechanisms enable parameter-efficient generalization in functions with low sensitivity.
2502.07555
O1 Embedder: Let Retrievers Think Before Action
cs.CL
The growing power of large language models (LLMs) has revolutionized how people access and utilize information. Notably, the LLMs excel at performing fine-grained data representation, which facilitates precise retrieval of information. They also generate high-quality answers based on external references, enabling the production of useful knowledge. The recent introduction of reasoning models, like OpenAI O1 and DeepSeek R1, marks another leap forward, highlighting LLMs' ability to think progressively before delivering final answers. This breakthrough significantly improves the ability to address complex tasks, e.g., coding and math proofs. Inspired by this progress, we aim to develop similar capabilities for retrieval models, which hold great promise for tackling critical challenges in the field, including multi-task retrieval, zero-shot retrieval, and tasks requiring intensive reasoning of complex relationships. With this motivation, we propose a novel approach called O1 Embedder, which generates useful thoughts for the input query before making retrieval for the target documents. To realize this objective, we conquer two technical difficulties. First, we design a data synthesis workflow, creating training signals for O1 Embedder by generating initial thoughts from an LLM-expert and subsequently refining them using a retrieval committee. Second, we optimize the training process, enabling a pre-trained model to be jointly fine-tuned to generate retrieval thoughts via behavior cloning and perform dense retrieval through contrastive learning. Our approach is evaluated by comprehensive experiments, where substantial improvements are achieved across 12 popular datasets, spanning both in-domain and out-of-domain scenarios. These results highlight O1 Embedder's remarkable accuracy and generalizability, paving the way for the development of next-generation IR foundation models.
2502.07556
SketchFlex: Facilitating Spatial-Semantic Coherence in Text-to-Image Generation with Region-Based Sketches
cs.HC cs.CV
Text-to-image models can generate visually appealing images from text descriptions. Efforts have been devoted to improving model controls with prompt tuning and spatial conditioning. However, our formative study highlights the challenges for non-expert users in crafting appropriate prompts and specifying fine-grained spatial conditions (e.g., depth or canny references) to generate semantically cohesive images, especially when multiple objects are involved. In response, we introduce SketchFlex, an interactive system designed to improve the flexibility of spatially conditioned image generation using rough region sketches. The system automatically infers user prompts with rational descriptions within a semantic space enriched by crowd-sourced object attributes and relationships. Additionally, SketchFlex refines users' rough sketches into canny-based shape anchors, ensuring the generation quality and alignment of user intentions. Experimental results demonstrate that SketchFlex achieves more cohesive image generations than end-to-end models, meanwhile significantly reducing cognitive load and better matching user intentions compared to region-based generation baseline.
2502.07558
Efficient Sparsification of Simplicial Complexes via Local Densities of States
stat.ML cs.CG cs.DM cs.NA cs.SI math.NA
Simplicial complexes (SCs), a generalization of graph models for relational data that account for higher-order relations between data items, have become a popular abstraction for analyzing complex data using tools from topological data analysis or topological signal processing. However, the analysis of many real-world datasets leads to dense SCs with a large number of higher-order interactions. Unfortunately, analyzing such large SCs often has a prohibitive cost in terms of computation time and memory consumption. The sparsification of such complexes, i.e., the approximation of an original SC with a sparser simplicial complex with only a log-linear number of high-order simplices while maintaining a spectrum close to the original SC, is of broad interest. In this work, we develop a novel method for a probabilistic sparsifaction of SCs. At its core lies the efficient computation of sparsifying sampling probability through local densities of states as functional descriptors of the spectral information. To avoid pathological structures in the spectrum of the corresponding Hodge Laplacian operators, we suggest a "kernel-ignoring" decomposition for approximating the sampling probability; additionally, we exploit error estimates to show asymptotically prevailing algorithmic complexity of the developed method. The performance of the framework is demonstrated on the family of Vietoris--Rips filtered simplicial complexes.
2502.07560
Navigating Semantic Drift in Task-Agnostic Class-Incremental Learning
cs.CV
Class-incremental learning (CIL) seeks to enable a model to sequentially learn new classes while retaining knowledge of previously learned ones. Balancing flexibility and stability remains a significant challenge, particularly when the task ID is unknown. To address this, our study reveals that the gap in feature distribution between novel and existing tasks is primarily driven by differences in mean and covariance moments. Building on this insight, we propose a novel semantic drift calibration method that incorporates mean shift compensation and covariance calibration. Specifically, we calculate each class's mean by averaging its sample embeddings and estimate task shifts using weighted embedding changes based on their proximity to the previous mean, effectively capturing mean shifts for all learned classes with each new task. We also apply Mahalanobis distance constraint for covariance calibration, aligning class-specific embedding covariances between old and current networks to mitigate the covariance shift. Additionally, we integrate a feature-level self-distillation approach to enhance generalization. Comprehensive experiments on commonly used datasets demonstrate the effectiveness of our approach. The source code is available at \href{https://github.com/fwu11/MACIL.git}{https://github.com/fwu11/MACIL.git}.
2502.07562
LoRP-TTS: Low-Rank Personalized Text-To-Speech
cs.SD cs.AI eess.AS
Speech synthesis models convert written text into natural-sounding audio. While earlier models were limited to a single speaker, recent advancements have led to the development of zero-shot systems that generate realistic speech from a wide range of speakers using their voices as additional prompts. However, they still struggle with imitating non-studio-quality samples that differ significantly from the training datasets. In this work, we demonstrate that utilizing Low-Rank Adaptation (LoRA) allows us to successfully use even single recordings of spontaneous speech in noisy environments as prompts. This approach enhances speaker similarity by up to $30pp$ while preserving content and naturalness. It represents a significant step toward creating truly diverse speech corpora, that is crucial in all speech-related tasks.
2502.07563
LASP-2: Rethinking Sequence Parallelism for Linear Attention and Its Hybrid
cs.LG cs.AI cs.CL
Linear sequence modeling approaches, such as linear attention, provide advantages like linear-time training and constant-memory inference over sequence lengths. However, existing sequence parallelism (SP) methods are either not optimized for the right-product-first feature of linear attention or use a ring-style communication strategy, which results in lower computation parallelism, limits their scalability for longer sequences in distributed systems. In this paper, we introduce LASP-2, a new SP method to enhance both communication and computation parallelism when training linear attention transformer models with very-long input sequences. Compared to previous work LASP, LASP-2 rethinks the minimal communication requirement for SP on linear attention layers, reorganizes the whole communication-computation workflow of LASP. In this way, only one single AllGather collective communication is needed on intermediate memory states, whose sizes are independent of the sequence length, leading to significant improvements of both communication and computation parallelism, as well as their overlap. Additionally, we extend LASP-2 to LASP-2H by applying similar communication redesign to standard attention modules, offering an efficient SP solution for hybrid models that blend linear and standard attention layers. Our evaluation on a Linear-Llama3 model, a variant of Llama3 with linear attention replacing standard attention, demonstrates the effectiveness of LASP-2 and LASP-2H. Specifically, LASP-2 achieves training speed improvements of 15.2% over LASP and 36.6% over Ring Attention, with a sequence length of 2048K across 64 GPUs. The Code is released as a part of: https://github.com/OpenSparseLLMs/Linear-MoE.
2502.07564
An Elliptic Curve Based Solution to the Perspective-Three-Point Problem
cs.CV math.AG
The Perspective-Three-Point Problem (P3P) is solved by first focusing on determining the directions of the lines through pairs of control points, relative to the camera, rather than the distances from the camera to the control points. The analysis of this produces an efficient, accurate and reasonably simple P3P solver, which is compared with a state-of-the-art P3P solver, "Lambda Twist." Both methods depend on the accurate computation of a single root of a cubic polynomial. They have been implemented and tested for a wide range of control-point triangles, and under certain reasonable restrictions, the new method is noticably more accurate than Lambda Twist, though it is slower. However, the principal value of the present work is not in introducing yet another P3P solver, but lies rather in the discovery of an intimate connection between the P3P problem and a special family of elliptic curves that includes curves utilized in cryptography. This holds the potential for further advances in a number of directions. To make this connection, an interesting spherical analogue of an ancient "sliding" problem is stated and solved.
2502.07566
Capacity of the Binary Energy Harvesting Channel
cs.IT math.IT
The capacity of a channel with an energy-harvesting (EH) encoder and a finite battery remains an open problem, even in the noiseless case. A key instance of this scenario is the binary EH channel (BEHC), where the encoder has a unit-sized battery and binary inputs. Existing capacity expressions for the BEHC are not computable, motivating this work, which determines the capacity to any desired precision via convex optimization. By modeling the system as a finite-state channel with state information known causally at the encoder, we derive single-letter lower and upper bounds using auxiliary directed graphs, termed $Q$-graphs. These $Q$-graphs exhibit a special structure with a finite number of nodes, $N$, enabling the formulation of the bounds as convex optimization problems. As $N$ increases, the bounds tighten and converge to the capacity with a vanishing gap of $O(N)$. For any EH probability parameter $\eta\in \{0.1,0.2, \dots, 0.9\}$, we compute the capacity with a precision of ${1e-6}$, outperforming the best-known bounds in the literature. Finally, we extend this framework to noisy EH channels with feedback, and present numerical achievable rates for the binary symmetric channel using a Markov decision process.
2502.07575
Towards Efficient and Multifaceted Computer-assisted Pronunciation Training Leveraging Hierarchical Selective State Space Model and Decoupled Cross-entropy Loss
eess.AS cs.CL
Prior efforts in building computer-assisted pronunciation training (CAPT) systems often treat automatic pronunciation assessment (APA) and mispronunciation detection and diagnosis (MDD) as separate fronts: the former aims to provide multiple pronunciation aspect scores across diverse linguistic levels, while the latter focuses instead on pinpointing the precise phonetic pronunciation errors made by non-native language learners. However, it is generally expected that a full-fledged CAPT system should perform both functionalities simultaneously and efficiently. In response to this surging demand, we in this work first propose HMamba, a novel CAPT approach that seamlessly integrates APA and MDD tasks in parallel. In addition, we introduce a novel loss function, decoupled cross-entropy loss (deXent), specifically tailored for MDD to facilitate better-supervised learning for detecting mispronounced phones, thereby enhancing overall performance. A comprehensive set of empirical results on the speechocean762 benchmark dataset demonstrates the effectiveness of our approach on APA. Notably, our proposed approach also yields a considerable improvement in MDD performance over a strong baseline, achieving an F1-score of 63.85%. Our codes are made available at https://github.com/Fuann/hmamba
2502.07577
Automated Capability Discovery via Model Self-Exploration
cs.LG cs.AI cs.CL
Foundation models have become general-purpose assistants, exhibiting diverse capabilities across numerous domains through training on web-scale data. It remains challenging to precisely characterize even a fraction of the full spectrum of capabilities and potential risks in any new model. Existing evaluation approaches often require significant human effort, and it is taking increasing effort to design ever harder challenges for more capable models. We introduce Automated Capability Discovery (ACD), a framework that designates one foundation model as a scientist to systematically propose open-ended tasks probing the abilities of a subject model (potentially itself). By combining frontier models with ideas from the field of open-endedness, ACD automatically and systematically uncovers both surprising capabilities and failures in the subject model. We demonstrate ACD across a range of foundation models (including the GPT, Claude, and Llama series), showing that it automatically reveals thousands of capabilities that would be challenging for any single team to uncover. We further validate our method's automated scoring with extensive human surveys, observing high agreement between model-generated and human evaluations. By leveraging foundation models' ability to both create tasks and self-evaluate, ACD is a significant step toward scalable, automated evaluation of novel AI systems. All code and evaluation logs are open-sourced at https://github.com/conglu1997/ACD.
2502.07579
Single-Step Consistent Diffusion Samplers
cs.LG stat.ML
Sampling from unnormalized target distributions is a fundamental yet challenging task in machine learning and statistics. Existing sampling algorithms typically require many iterative steps to produce high-quality samples, leading to high computational costs that limit their practicality in time-sensitive or resource-constrained settings. In this work, we introduce consistent diffusion samplers, a new class of samplers designed to generate high-fidelity samples in a single step. We first develop a distillation algorithm to train a consistent diffusion sampler from a pretrained diffusion model without pre-collecting large datasets of samples. Our algorithm leverages incomplete sampling trajectories and noisy intermediate states directly from the diffusion process. We further propose a method to train a consistent diffusion sampler from scratch, fully amortizing exploration by training a single model that both performs diffusion sampling and skips intermediate steps using a self-consistency loss. Through extensive experiments on a variety of unnormalized distributions, we show that our approach yields high-fidelity samples using less than 1% of the network evaluations required by traditional diffusion samplers.
2502.07580
Generative Modeling with Bayesian Sample Inference
cs.LG stat.ML
We derive a novel generative model from the simple act of Gaussian posterior inference. Treating the generated sample as an unknown variable to infer lets us formulate the sampling process in the language of Bayesian probability. Our model uses a sequence of prediction and posterior update steps to narrow down the unknown sample from a broad initial belief. In addition to a rigorous theoretical analysis, we establish a connection between our model and diffusion models and show that it includes Bayesian Flow Networks (BFNs) as a special case. In our experiments, we demonstrate improved performance over both BFNs and Variational Diffusion Models, achieving competitive likelihood scores on CIFAR10 and ImageNet.
2502.07584
Understanding the Generalization Error of Markov algorithms through Poissonization
stat.ML cs.LG
Using continuous-time stochastic differential equation (SDE) proxies to stochastic optimization algorithms has proven fruitful for understanding their generalization abilities. A significant part of these approaches are based on the so-called ``entropy flows'', which greatly simplify the generalization analysis. Unfortunately, such well-structured entropy flows cannot be obtained for most discrete-time algorithms, and the existing SDE approaches remain limited to specific noise and algorithmic structures. We aim to alleviate this issue by introducing a generic framework for analyzing the generalization error of Markov algorithms through `Poissonization', a continuous-time approximation of discrete-time processes with formal approximation guarantees. Through this approach, we first develop a novel entropy flow, which directly leads to PAC-Bayesian generalization bounds. We then draw novel links to modified versions of the celebrated logarithmic Sobolev inequalities (LSI), identify cases where such LSIs are satisfied, and obtain improved bounds. Beyond its generality, our framework allows exploiting specific properties of learning algorithms. In particular, we incorporate the noise structure of different algorithm types - namely, those with additional noise injections (noisy) and those without (non-noisy) - through various technical tools. This illustrates the capacity of our methods to achieve known (yet, Poissonized) and new generalization bounds.
2502.07586
We Can't Understand AI Using our Existing Vocabulary
cs.CL cs.AI
This position paper argues that, in order to understand AI, we cannot rely on our existing vocabulary of human words. Instead, we should strive to develop neologisms: new words that represent precise human concepts that we want to teach machines, or machine concepts that we need to learn. We start from the premise that humans and machines have differing concepts. This means interpretability can be framed as a communication problem: humans must be able to reference and control machine concepts, and communicate human concepts to machines. Creating a shared human-machine language through developing neologisms, we believe, could solve this communication problem. Successful neologisms achieve a useful amount of abstraction: not too detailed, so they're reusable in many contexts, and not too high-level, so they convey precise information. As a proof of concept, we demonstrate how a "length neologism" enables controlling LLM response length, while a "diversity neologism" allows sampling more variable responses. Taken together, we argue that we cannot understand AI using our existing vocabulary, and expanding it through neologisms creates opportunities for both controlling and understanding machines better.
2502.07587
SEMU: Singular Value Decomposition for Efficient Machine Unlearning
cs.LG
While the capabilities of generative foundational models have advanced rapidly in recent years, methods to prevent harmful and unsafe behaviors remain underdeveloped. Among the pressing challenges in AI safety, machine unlearning (MU) has become increasingly critical to meet upcoming safety regulations. Most existing MU approaches focus on altering the most significant parameters of the model. However, these methods often require fine-tuning substantial portions of the model, resulting in high computational costs and training instabilities, which are typically mitigated by access to the original training dataset. In this work, we address these limitations by leveraging Singular Value Decomposition (SVD) to create a compact, low-dimensional projection that enables the selective forgetting of specific data points. We propose Singular Value Decomposition for Efficient Machine Unlearning (SEMU), a novel approach designed to optimize MU in two key aspects. First, SEMU minimizes the number of model parameters that need to be modified, effectively removing unwanted knowledge while making only minimal changes to the model's weights. Second, SEMU eliminates the dependency on the original training dataset, preserving the model's previously acquired knowledge without additional data requirements. Extensive experiments demonstrate that SEMU achieves competitive performance while significantly improving efficiency in terms of both data usage and the number of modified parameters.
2502.07590
DSV: Exploiting Dynamic Sparsity to Accelerate Large-Scale Video DiT Training
cs.DC cs.CV
Diffusion Transformers (DiTs) have shown remarkable performance in modeling and generating high-quality videos. However, the quadratic computational complexity of 3D full attention mechanism presents significant challenges in scaling video DiT training, especially for high-definition and lengthy videos, where attention can dominate up to 95% of the end-to-end time and necessitate specialized communication paradigms to handle large input sizes. This paper introduces DSV, a novel framework designed to accelerate and scale the training of video DiTs by leveraging the inherent dynamic attention sparsity throughout the training process. DSV employs a two-stage training algorithm that exploits sparsity patterns, focusing on critical elements supported by efficient, tailored kernels. To accommodate the new sparsity dimension, we develop a hybrid sparsity-aware context parallelism that effectively scales to large inputs by addressing the heterogeneity of sparsity across attention heads and blocks, resulting in optimized sparse computation and communication. Extensive evaluations demonstrate that DSV achieves up to 3.02x gain in training throughput with nearly no quality degradation.
2502.07591
DMWM: Dual-Mind World Model with Long-Term Imagination
cs.LG cs.AI
Imagination in world models is crucial for enabling agents to learn long-horizon policy in a sample-efficient manner. Existing recurrent state-space model (RSSM)-based world models depend on single-step statistical inference to capture the environment dynamics, and, hence, they are unable to perform long-term imagination tasks due to the accumulation of prediction errors. Inspired by the dual-process theory of human cognition, we propose a novel dual-mind world model (DMWM) framework that integrates logical reasoning to enable imagination with logical consistency. DMWM is composed of two components: an RSSM-based System 1 (RSSM-S1) component that handles state transitions in an intuitive manner and a logic-integrated neural network-based System 2 (LINN-S2) component that guides the imagination process through hierarchical deep logical reasoning. The inter-system feedback mechanism is designed to ensure that the imagination process follows the logical rules of the real environment. The proposed framework is evaluated on benchmark tasks that require long-term planning from the DMControl suite. Extensive experimental results demonstrate that the proposed framework yields significant improvements in terms of logical coherence, trial efficiency, data efficiency and long-term imagination over the state-of-the-art world models.
2502.07592
YOLO Network For Defect Detection In Optical lenses
cs.CV
Mass-produced optical lenses often exhibit defects that alter their scattering properties and compromise quality standards. Manual inspection is usually adopted to detect defects, but it is not recommended due to low accuracy, high error rate and limited scalability. To address these challenges, this study presents an automated defect detection system based on the YOLOv8 deep learning model. A custom dataset of optical lenses, annotated with defect and lens regions, was created to train the model. Experimental results obtained in this study reveal that the system can be used to efficiently and accurately detect defects in optical lenses. The proposed system can be utilized in real-time industrial environments to enhance quality control processes by enabling reliable and scalable defect detection in optical lens manufacturing.
2502.07595
Distributed Coverage Control for Time-Varying Spatial Processes
cs.RO
Multi-robot systems are essential for environmental monitoring, particularly for tracking spatial phenomena like pollution, soil minerals, and water salinity, and more. This study addresses the challenge of deploying a multi-robot team for optimal coverage in environments where the density distribution, describing areas of interest, is unknown and changes over time. We propose a fully distributed control strategy that uses Gaussian Processes (GPs) to model the spatial field and balance the trade-off between learning the field and optimally covering it. Unlike existing approaches, we address a more realistic scenario by handling time-varying spatial fields, where the exploration-exploitation trade-off is dynamically adjusted over time. Each robot operates locally, using only its own collected data and the information shared by the neighboring robots. To address the computational limits of GPs, the algorithm efficiently manages the volume of data by selecting only the most relevant samples for the process estimation. The performance of the proposed algorithm is evaluated through several simulations and experiments, incorporating real-world data phenomena to validate its effectiveness.
2502.07596
Enviro-IoT: Calibrating Low-Cost Environmental Sensors in Urban Settings
eess.SY cs.SY
Low-cost miniaturised sensors offer significant advantage to monitor the environment in real-time and accurately. The area of air quality monitoring has attracted much attention in recent years because of the increasing impacts on the environment and more personally to human health and mental wellbeing. Rapid growth in sensors and Internet of Things (IoT) technologies is paving the way for low-cost systems to transform global monitoring of air quality. Drawing on 4 years of development work, in this paper we outline the design, implementation and analysis of \textit{Enviro-IoT} as a step forward to monitoring air quality levels within urban environments by means of a low-cost sensing system. An in-the-wild study for 9-months was performed to evaluate the Enviro-IoT system against industry standard equipment is performed with accuracy for measuring Particulate Matter 2.5, 10 and Nitrogen Dioxide achieving 98\%, 97\% and 97\% respectively. The results in this case study are made up Of 57, 120 which highlight that it is possible to take advantage of low-cost sensors coupled with IoT technologies to validate the Enviro-IoT device against research-grade industrial instruments.
2502.07599
DPO-Shift: Shifting the Distribution of Direct Preference Optimization
cs.CL
Direct Preference Optimization (DPO) and its variants have become increasingly popular for aligning language models with human preferences. These methods aim to teach models to better distinguish between chosen (or preferred) and rejected (or dispreferred) responses. However, prior research has identified that the probability of chosen responses often decreases during training, and this phenomenon is known as likelihood displacement. To tackle this challenge, in this work we introduce \method to controllably shift the distribution of the chosen probability. Then, we show that \method exhibits a fundamental trade-off between improving the chosen probability and sacrificing the reward margin, as supported by both theoretical analysis and experimental validation. Furthermore, we demonstrate the superiority of \method over DPO on downstream tasks such as MT-Bench and a designed win rate experiment. We believe this study shows that the likelihood displacement issue of DPO can be effectively mitigated with a simple, theoretically grounded solution. Our code is available at https://github.com/Meaquadddd/DPO-Shift.
2502.07600
PlaySlot: Learning Inverse Latent Dynamics for Controllable Object-Centric Video Prediction and Planning
cs.CV cs.RO
Predicting future scene representations is a crucial task for enabling robots to understand and interact with the environment. However, most existing methods rely on video sequences and simulations with precise action annotations, limiting their ability to leverage the large amount of available unlabeled video data. To address this challenge, we propose PlaySlot, an object-centric video prediction model that infers object representations and latent actions from unlabeled video sequences. It then uses these representations to forecast future object states and video frames. PlaySlot allows to generate multiple possible futures conditioned on latent actions, which can be inferred from video dynamics, provided by a user, or generated by a learned action policy, thus enabling versatile and interpretable world modeling. Our results show that PlaySlot outperforms both stochastic and object-centric baselines for video prediction across different environments. Furthermore, we show that our inferred latent actions can be used to learn robot behaviors sample-efficiently from unlabeled video demonstrations. Videos and code are available at https://play-slot.github.io/PlaySlot/.
2502.07601
Towards Zero-Shot Anomaly Detection and Reasoning with Multimodal Large Language Models
cs.CV cs.CL
Zero-Shot Anomaly Detection (ZSAD) is an emerging AD paradigm. Unlike the traditional unsupervised AD setting that requires a large number of normal samples to train a model, ZSAD is more practical for handling data-restricted real-world scenarios. Recently, Multimodal Large Language Models (MLLMs) have shown revolutionary reasoning capabilities in various vision tasks. However, the reasoning of image abnormalities remains underexplored due to the lack of corresponding datasets and benchmarks. To facilitate research in AD & reasoning, we establish the first visual instruction tuning dataset, Anomaly-Instruct-125k, and the evaluation benchmark, VisA-D&R. Through investigation with our benchmark, we reveal that current MLLMs like GPT-4o cannot accurately detect and describe fine-grained anomalous details in images. To address this, we propose Anomaly-OneVision (Anomaly-OV), the first specialist visual assistant for ZSAD and reasoning. Inspired by human behavior in visual inspection, Anomaly-OV leverages a Look-Twice Feature Matching (LTFM) mechanism to adaptively select and emphasize abnormal visual tokens. Extensive experiments demonstrate that Anomaly-OV achieves significant improvements over advanced generalist models in both detection and reasoning. Extensions to medical and 3D AD are provided for future study. The link to our project page: https://xujiacong.github.io/Anomaly-OV/
2502.07602
An Improved Optimal Proximal Gradient Algorithm for Non-Blind Image Deblurring
cs.CV math.OC
Image deblurring remains a central research area within image processing, critical for its role in enhancing image quality and facilitating clearer visual representations across diverse applications. This paper tackles the optimization problem of image deblurring, assuming a known blurring kernel. We introduce an improved optimal proximal gradient algorithm (IOptISTA), which builds upon the optimal gradient method and a weighting matrix, to efficiently address the non-blind image deblurring problem. Based on two regularization cases, namely the $l_1$ norm and total variation norm, we perform numerical experiments to assess the performance of our proposed algorithm. The results indicate that our algorithm yields enhanced PSNR and SSIM values, as well as a reduced tolerance, compared to existing methods.
2502.07606
Algorithmic Aspects of Strategic Trading
cs.GT cs.CE cs.LG
Algorithmic trading in modern financial markets is widely acknowledged to exhibit strategic, game-theoretic behaviors whose complexity can be difficult to model. A recent series of papers (Chriss, 2024b,c,a, 2025) has made progress in the setting of trading for position building. Here parties wish to buy or sell a fixed number of shares in a fixed time period in the presence of both temporary and permanent market impact, resulting in exponentially large strategy spaces. While these papers primarily consider the existence and structural properties of equilibrium strategies, in this work we focus on the algorithmic aspects of the proposed model. We give an efficient algorithm for computing best responses, and show that while the temporary impact only setting yields a potential game, best response dynamics do not generally converge for the general setting, for which no fast algorithm for (Nash) equilibrium computation is known. This leads us to consider the broader notion of Coarse Correlated Equilibria (CCE), which we show can be computed efficiently via an implementation of Follow the Perturbed Leader (FTPL). We illustrate the model and our results with an experimental investigation, where FTPL exhibits interesting behavior in different regimes of the relative weighting between temporary and permanent market impact.
2502.07608
Beyond Prompting: Time2Lang -- Bridging Time-Series Foundation Models and Large Language Models for Health Sensing
cs.LG cs.HC
Large language models (LLMs) show promise for health applications when combined with behavioral sensing data. Traditional approaches convert sensor data into text prompts, but this process is prone to errors, computationally expensive, and requires domain expertise. These challenges are particularly acute when processing extended time series data. While time series foundation models (TFMs) have recently emerged as powerful tools for learning representations from temporal data, bridging TFMs and LLMs remains challenging. Here, we present Time2Lang, a framework that directly maps TFM outputs to LLM representations without intermediate text conversion. Our approach first trains on synthetic data using periodicity prediction as a pretext task, followed by evaluation on mental health classification tasks. We validate Time2Lang on two longitudinal wearable and mobile sensing datasets: daily depression prediction using step count data (17,251 days from 256 participants) and flourishing classification based on conversation duration (46 participants over 10 weeks). Time2Lang maintains near constant inference times regardless of input length, unlike traditional prompting methods. The generated embeddings preserve essential time-series characteristics such as auto-correlation. Our results demonstrate that TFMs and LLMs can be effectively integrated while minimizing information loss and enabling performance transfer across these distinct modeling paradigms. To our knowledge, we are the first to integrate a TFM and an LLM for health, thus establishing a foundation for future research combining general-purpose large models for complex healthcare tasks.
2502.07615
Flow Distillation Sampling: Regularizing 3D Gaussians with Pre-trained Matching Priors
cs.CV
3D Gaussian Splatting (3DGS) has achieved excellent rendering quality with fast training and rendering speed. However, its optimization process lacks explicit geometric constraints, leading to suboptimal geometric reconstruction in regions with sparse or no observational input views. In this work, we try to mitigate the issue by incorporating a pre-trained matching prior to the 3DGS optimization process. We introduce Flow Distillation Sampling (FDS), a technique that leverages pre-trained geometric knowledge to bolster the accuracy of the Gaussian radiance field. Our method employs a strategic sampling technique to target unobserved views adjacent to the input views, utilizing the optical flow calculated from the matching model (Prior Flow) to guide the flow analytically calculated from the 3DGS geometry (Radiance Flow). Comprehensive experiments in depth rendering, mesh reconstruction, and novel view synthesis showcase the significant advantages of FDS over state-of-the-art methods. Additionally, our interpretive experiments and analysis aim to shed light on the effects of FDS on geometric accuracy and rendering quality, potentially providing readers with insights into its performance. Project page: https://nju-3dv.github.io/projects/fds
2502.07616
Tractable Transformers for Flexible Conditional Generation
cs.CL cs.LG
Non-autoregressive (NAR) generative models are valuable because they can handle diverse conditional generation tasks in a more principled way than their autoregressive (AR) counterparts, which are constrained by sequential dependency requirements. Recent advancements in NAR models, such as diffusion language models, have demonstrated superior performance in unconditional generation compared to AR models (e.g., GPTs) of similar sizes. However, such improvements do not always lead to improved conditional generation performance. We show that a key reason for this gap is the difficulty in generalizing to conditional probability queries unseen during training. As a result, strong unconditional generation performance does not guarantee high-quality conditional generation. This paper proposes Tractable Transformers (Tracformer), a Transformer-based generative model that is more robust to different conditional generation tasks. Unlike existing models that rely solely on global contextual features derived from full inputs, Tracformers incorporate a sparse Transformer encoder to capture both local and global contextual information. This information is routed through a decoder for conditional generation. Empirical results demonstrate that Tracformers achieve state-of-the-art conditional generation performance on text modeling compared to recent diffusion and AR model baselines.
2502.07617
Scaling Pre-training to One Hundred Billion Data for Vision Language Models
cs.CV
We provide an empirical investigation of the potential of pre-training vision-language models on an unprecedented scale: 100 billion examples. We find that model performance tends to saturate at this scale on many common Western-centric classification and retrieval benchmarks, such as COCO Captions. Nevertheless, tasks of cultural diversity achieve more substantial gains from the 100-billion scale web data, thanks to its coverage of long-tail concepts. Furthermore, we analyze the model's multilinguality and show gains in low-resource languages as well. In addition, we observe that reducing the size of the pretraining dataset via quality filters like using CLIP, typically used to enhance performance, may inadvertently reduce the cultural diversity represented even in large-scale datasets. Our results highlight that while traditional benchmarks may not benefit significantly from scaling noisy, raw web data to 100 billion examples, this data scale is vital for building truly inclusive multimodal systems.
2502.07620
Causal-Informed Contrastive Learning: Towards Bias-Resilient Pre-training under Concept Drift
cs.LG cs.CV
The evolution of large-scale contrastive pre-training propelled by top-tier datasets has reached a transition point in the scaling law. Consequently, sustaining and enhancing a model's pre-training capabilities in drift environments have surfaced as a notable challenge. In this paper, we initially uncover that contrastive pre-training methods are significantly impacted by concept drift wherein distributions change unpredictably, resulting in notable biases in the feature space of the pre-trained model. Empowered by causal inference, we construct a structural causal graph to analyze the impact of concept drift to contrastive pre-training systemically, and propose the causal interventional contrastive objective. Upon achieving this, we devise a resilient contrastive pre-training approach to accommodate the data stream of concept drift, with simple and scalable implementation. Extensive experiments on various downstream tasks demonstrate our resilient contrastive pre-training effectively mitigates the bias stemming from the concept drift data stream. Codes are available at https://anonymous.4open.science/r/ResilientCL/.
2502.07623
Lexical categories of stem-forming roots in Mapud\"ungun verb forms
cs.CL
After developing a computational system for morphological analysis of the Mapuche language, and evaluating it with texts from various authors and styles, it became necessary to verify the linguistic assumptions of the source used as the basis for implementing this tool. In the present work, the primary focus is on the lexical category classification of Mapud\"ungun roots recognised as verbal in the source utilised for the development of the morphological analysis system. The results of this lexical category revision directly benefit the computational analyser, as they are implemented as soon as they are verified. Additionally, it is hoped that these results will help clarify some uncertainties about lexical categories in the Mapuche language. This work addresses a preliminary task to identify the valency of true verbal roots, the results of which will be presented in a subsequent work that complements this article.
2502.07629
Exploring Mobile Touch Interaction with Large Language Models
cs.HC cs.CL
Interacting with Large Language Models (LLMs) for text editing on mobile devices currently requires users to break out of their writing environment and switch to a conversational AI interface. In this paper, we propose to control the LLM via touch gestures performed directly on the text. We first chart a design space that covers fundamental touch input and text transformations. In this space, we then concretely explore two control mappings: spread-to-generate and pinch-to-shorten, with visual feedback loops. We evaluate this concept in a user study (N=14) that compares three feedback designs: no visualisation, text length indicator, and length + word indicator. The results demonstrate that touch-based control of LLMs is both feasible and user-friendly, with the length + word indicator proving most effective for managing text generation. This work lays the foundation for further research into gesture-based interaction with LLMs on touch devices.
2502.07630
Rethinking Timing Residuals: Advancing PET Detectors with Explicit TOF Corrections
physics.ins-det cs.LG
PET is a functional imaging method that visualizes metabolic processes. TOF information can be derived from coincident detector signals and incorporated into image reconstruction to enhance the SNR. PET detectors are typically assessed by their CTR, but timing performance is degraded by various factors. Research on timing calibration seeks to mitigate these degradations and restore accurate timing information. While many calibration methods use analytical approaches, machine learning techniques have recently gained attention due to their flexibility. We developed a residual physics-based calibration approach that combines prior domain knowledge with the power of machine learning models. This approach begins with an initial analytical calibration addressing first-order skews. The remaining deviations, regarded as residual effects, are used to train machine learning models to eliminate higher-order skews. The key advantage is that the experimenter guides the learning process through the definition of timing residuals. In earlier studies, we developed models that directly predicted the expected time difference, which offered corrections only implicitly (implicit correction models). In this study, we introduce a new definition for timing residuals, enabling us to train models that directly predict correction values (explicit correction models). The explicit correction approach significantly simplifies data acquisition, improves linearity, and enhances timing performance from $371 \pm 6$ ps to $281 \pm 5$ ps for coincidences from 430 keV to 590 keV. Additionally, the new definition reduces model size, making it suitable for high-throughput applications like PET scanners. Experiments were conducted using two detector stacks composed of $4 \times 4$ LYSO:Ce,Ca crystals ($3.8\times 3.8\times 20$ mm$^{3}$) coupled to $4 \times 4$ Broadcom NUV-MT SiPMs and digitized with the TOFPET2 ASIC.
2502.07631
Divide and Merge: Motion and Semantic Learning in End-to-End Autonomous Driving
cs.CV
Perceiving the environment and its changes over time corresponds to two fundamental yet heterogeneous types of information: semantics and motion. Previous end-to-end autonomous driving works represent both types of information in a single feature vector. However, including motion tasks, such as prediction and planning, always impairs detection and tracking performance, a phenomenon known as negative transfer in multi-task learning. To address this issue, we propose Neural-Bayes motion decoding, a novel parallel detection, tracking, and prediction method separating semantic and motion learning, similar to the Bayes filter. Specifically, we employ a set of learned motion queries that operate in parallel with the detection and tracking queries, sharing a unified set of recursively updated reference points. Moreover, we employ interactive semantic decoding to enhance information exchange in semantic tasks, promoting positive transfer. Experiments on the nuScenes dataset show improvements of 5% in detection and 11% in tracking. Our method achieves state-of-the-art collision rates in open-loop planning evaluation without any modifications to the planning module.
2502.07634
Efficient Distributed Training through Gradient Compression with Sparsification and Quantization Techniques
cs.LG cs.MM
This study investigates the impact of gradient compression on distributed training performance, focusing on sparsification and quantization techniques, including top-k, DGC, and QSGD. In baseline experiments, random-k compression results in severe performance degradation, highlighting its inefficacy. In contrast, using top-k and DGC at 50 times compression yields performance improvements, reducing perplexity by up to 0.06 compared to baseline. Experiments across 1, 2, and 4 workers demonstrate that conservative sparsification can have a regularizing effect, especially for smaller models, while compression ratios above 5000 times impair performance, particularly for DGC. Communication times are reduced across all compression methods, with top-k and DGC decreasing communication to negligible levels at high compression ratios. However, increased computation times offset this efficiency for top-k due to sorting demands, making it less scalable than DGC or QSGD. In convergence tests, sparsification techniques show accelerated convergence, requiring fewer epochs than the baseline, which has implications for computational savings. Although precision trade-offs emerge, floating point errors are mitigated by compression. This study's findings underscore the need to tune hyperparameters specifically for each compression technique to achieve optimal model performance, especially in distributed training systems.
2502.07635
Distributed Value Decomposition Networks with Networked Agents
cs.LG cs.AI cs.MA
We investigate the problem of distributed training under partial observability, whereby cooperative multi-agent reinforcement learning agents (MARL) maximize the expected cumulative joint reward. We propose distributed value decomposition networks (DVDN) that generate a joint Q-function that factorizes into agent-wise Q-functions. Whereas the original value decomposition networks rely on centralized training, our approach is suitable for domains where centralized training is not possible and agents must learn by interacting with the physical environment in a decentralized manner while communicating with their peers. DVDN overcomes the need for centralized training by locally estimating the shared objective. We contribute with two innovative algorithms, DVDN and DVDN (GT), for the heterogeneous and homogeneous agents settings respectively. Empirically, both algorithms approximate the performance of value decomposition networks, in spite of the information loss during communication, as demonstrated in ten MARL tasks in three standard environments.
2502.07636
Consistency Training with Physical Constraints
cs.LG
We propose a physics-aware Consistency Training (CT) method that accelerates sampling in Diffusion Models with physical constraints. Our approach leverages a two-stage strategy: (1) learning the noise-to-data mapping via CT, and (2) incorporating physics constraints as a regularizer. Experiments on toy examples show that our method generates samples in a single step while adhering to the imposed constraints. This approach has the potential to efficiently solve partial differential equations (PDEs) using deep generative modeling.
2502.07637
BiaSWE: An Expert Annotated Dataset for Misogyny Detection in Swedish
cs.CL
In this study, we introduce the process for creating BiaSWE, an expert-annotated dataset tailored for misogyny detection in the Swedish language. To address the cultural and linguistic specificity of misogyny in Swedish, we collaborated with experts from the social sciences and humanities. Our interdisciplinary team developed a rigorous annotation process, incorporating both domain knowledge and language expertise, to capture the nuances of misogyny in a Swedish context. This methodology ensures that the dataset is not only culturally relevant but also aligned with broader efforts in bias detection for low-resource languages. The dataset, along with the annotation guidelines, is publicly available for further research.
2502.07640
Goedel-Prover: A Frontier Model for Open-Source Automated Theorem Proving
cs.LG cs.AI
We introduce Goedel-Prover, an open-source large language model (LLM) that achieves the state-of-the-art (SOTA) performance in automated formal proof generation for mathematical problems. The key challenge in this field is the scarcity of formalized math statements and proofs, which we tackle in the following ways. We train statement formalizers to translate the natural language math problems from Numina into formal language (Lean 4), creating a dataset of 1.64 million formal statements. LLMs are used to check that the formal statements accurately preserve the content of the original natural language problems. We then iteratively build a large dataset of formal proofs by training a series of provers. Each prover succeeds in proving many statements that the previous ones could not, and these new proofs are added to the training set for the next prover. Despite using only supervised fine-tuning, our final prover significantly outperforms the previous best open-source model, DeepSeek-Prover-V1.5, which employs reinforcement learning. On the miniF2F benchmark, our model achieves a success rate of 57.6% (Pass@32), surpassing DeepSeek-Prover-V1.5 by 7.6%. On PutnamBench, Goedel-Prover successfully solves 7 problems (Pass@512), ranking first on the leaderboard. Furthermore, it generates 29.7K formal proofs for Lean Workbook problems, nearly doubling the 15.7K produced by earlier works.
2502.07642
FoQA: A Faroese Question-Answering Dataset
cs.CL cs.LG
We present FoQA, a Faroese extractive question-answering (QA) dataset with 2,000 samples, created using a semi-automated approach combining Large Language Models (LLMs) and human validation. The dataset was generated from Faroese Wikipedia articles using GPT-4-turbo for initial QA generation, followed by question rephrasing to increase complexity and native speaker validation to ensure quality. We provide baseline performance metrics for FoQA across multiple models, including LLMs and BERT, demonstrating its effectiveness in evaluating Faroese QA performance. The dataset is released in three versions: a validated set of 2,000 samples, a complete set of all 10,001 generated samples, and a set of 2,395 rejected samples for error analysis.
2502.07644
SymGPT: Auditing Smart Contracts via Combining Symbolic Execution with Large Language Models
cs.AI
To govern smart contracts running on Ethereum, multiple Ethereum Request for Comment (ERC) standards have been developed, each having a set of rules to guide the behaviors of smart contracts. Violating the ERC rules could cause serious security issues and financial loss, signifying the importance of verifying smart contracts follow ERCs. Today's practices of such verification are to manually audit each single contract, use expert-developed program-analysis tools, or use large language models (LLMs), all of which are far from effective in identifying ERC rule violations. This paper introduces SymGPT, a tool that combines the natural language understanding of large language models (LLMs) with the formal guarantees of symbolic execution to automatically verify smart contracts' compliance with ERC rules. To develop SymGPT, we conduct an empirical study of 132 ERC rules from three widely used ERC standards, examining their content, security implications, and natural language descriptions. Based on this study, we design SymGPT by first instructing an LLM to translate ERC rules into a defined EBNF grammar. We then synthesize constraints from the formalized rules to represent scenarios where violations may occur and use symbolic execution to detect them. Our evaluation shows that SymGPT identifies 5,783 ERC rule violations in 4,000 real-world contracts, including 1,375 violations with clear attack paths for stealing financial assets, demonstrating its effectiveness. Furthermore, SymGPT outperforms six automated techniques and a security-expert auditing service, underscoring its superiority over current smart contract analysis methods.