Search is not available for this dataset
id
string
submitter
string
authors
string
title
string
comments
string
journal-ref
string
doi
string
report-no
string
categories
string
license
string
abstract
string
versions
list
update_date
timestamp[s]
authors_parsed
list
prompt
string
2503.09793
Shailendra Joshi
Shailendra P. Joshi, Ashley Bucsek, Darren C. Pagan, Samantha Daly, Suraj Ravindran, Jaime Marian, Miguel A. Bessa, Surya R. Kalidindi, Nikhil C. Admal, Celia Reina, Somnath Ghosh, Jorge Vinals, and Ellad B.Tadmor
Integrated Experiment and Simulation Co-Design: A Key Infrastructure for Predictive Mesoscale Materials Modeling
null
null
null
null
cond-mat.mtrl-sci physics.comp-ph
http://creativecommons.org/licenses/by/4.0/
The design of structural & functional materials for specialized applications is being fueled by rapid advancements in materials synthesis, characterization, manufacturing, with sophisticated computational materials modeling frameworks that span a wide spectrum of length & time scales in the mesoscale between atomistic & continuum approaches. This is leading towards a systems-based design methodology that will replace traditional empirical approaches, embracing the principles of the Materials Genome Initiative. However, several gaps remain in this framework as it relates to advanced structural materials:(1) limited availability & access to high-fidelity experimental & computational datasets, (2) lack of co-design of experiments & simulation aimed at computational model validation,(3) lack of on-demand access to verified and validated codes for simulation and for experimental analyses, & (4) limited opportunities for workforce training and educational outreach. These shortcomings stifle major innovations in structural materials design. This paper describes plans for a community-driven research initiative that addresses current gaps based on best-practice recommendations of leaders in mesoscale modeling, experimentation & cyberinfrastructure obtained at an NSF-sponsored workshop dedicated to this topic. The proposal is to create a hub for Mesoscale Experimentation and Simulation co-Operation (hMESO)-that will (I) provide curation and sharing of models, data, & codes, (II) foster co-design of experiments for model validation with systematic uncertainty quantification, & (III) provide a platform for education & workforce development. It will engage experimental & computational experts in mesoscale mechanics and plasticity, along with mathematicians and computer scientists with expertise in algorithms, data science, machine learning, & large-scale cyberinfrastructure initiatives.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 19:55:34 GMT" } ]
2025-03-14T00:00:00
[ [ "Joshi", "Shailendra P.", "" ], [ "Bucsek", "Ashley", "" ], [ "Pagan", "Darren C.", "" ], [ "Daly", "Samantha", "" ], [ "Ravindran", "Suraj", "" ], [ "Marian", "Jaime", "" ], [ "Bessa", "Miguel A.", "" ],...
TITLE: Integrated Experiment and Simulation Co-Design: A Key Infrastructure for Predictive Mesoscale Materials Modeling ABSTRACT: The design of structural & functional materials for specialized applications is being fueled by rapid advancements in materials synthesis, characterization, manufacturing, with sophisticated computational materials modeling frameworks that span a wide spectrum of length & time scales in the mesoscale between atomistic & continuum approaches. This is leading towards a systems-based design methodology that will replace traditional empirical approaches, embracing the principles of the Materials Genome Initiative. However, several gaps remain in this framework as it relates to advanced structural materials:(1) limited availability & access to high-fidelity experimental & computational datasets, (2) lack of co-design of experiments & simulation aimed at computational model validation,(3) lack of on-demand access to verified and validated codes for simulation and for experimental analyses, & (4) limited opportunities for workforce training and educational outreach. These shortcomings stifle major innovations in structural materials design. This paper describes plans for a community-driven research initiative that addresses current gaps based on best-practice recommendations of leaders in mesoscale modeling, experimentation & cyberinfrastructure obtained at an NSF-sponsored workshop dedicated to this topic. The proposal is to create a hub for Mesoscale Experimentation and Simulation co-Operation (hMESO)-that will (I) provide curation and sharing of models, data, & codes, (II) foster co-design of experiments for model validation with systematic uncertainty quantification, & (III) provide a platform for education & workforce development. It will engage experimental & computational experts in mesoscale mechanics and plasticity, along with mathematicians and computer scientists with expertise in algorithms, data science, machine learning, & large-scale cyberinfrastructure initiatives.
2503.09797
Benjamin Towle
Benjamin Towle, Xin Chen, Ke Zhou
SeqSAM: Autoregressive Multiple Hypothesis Prediction for Medical Image Segmentation using SAM
Accepted to ISBI 2025
null
null
null
cs.CV cs.AI cs.LG
http://creativecommons.org/licenses/by/4.0/
Pre-trained segmentation models are a powerful and flexible tool for segmenting images. Recently, this trend has extended to medical imaging. Yet, often these methods only produce a single prediction for a given image, neglecting inherent uncertainty in medical images, due to unclear object boundaries and errors caused by the annotation tool. Multiple Choice Learning is a technique for generating multiple masks, through multiple learned prediction heads. However, this cannot readily be extended to producing more outputs than its initial pre-training hyperparameters, as the sparse, winner-takes-all loss function makes it easy for one prediction head to become overly dominant, thus not guaranteeing the clinical relevancy of each mask produced. We introduce SeqSAM, a sequential, RNN-inspired approach to generating multiple masks, which uses a bipartite matching loss for ensuring the clinical relevancy of each mask, and can produce an arbitrary number of masks. We show notable improvements in quality of each mask produced across two publicly available datasets. Our code is available at https://github.com/BenjaminTowle/SeqSAM.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 20:01:52 GMT" } ]
2025-03-14T00:00:00
[ [ "Towle", "Benjamin", "" ], [ "Chen", "Xin", "" ], [ "Zhou", "Ke", "" ] ]
TITLE: SeqSAM: Autoregressive Multiple Hypothesis Prediction for Medical Image Segmentation using SAM ABSTRACT: Pre-trained segmentation models are a powerful and flexible tool for segmenting images. Recently, this trend has extended to medical imaging. Yet, often these methods only produce a single prediction for a given image, neglecting inherent uncertainty in medical images, due to unclear object boundaries and errors caused by the annotation tool. Multiple Choice Learning is a technique for generating multiple masks, through multiple learned prediction heads. However, this cannot readily be extended to producing more outputs than its initial pre-training hyperparameters, as the sparse, winner-takes-all loss function makes it easy for one prediction head to become overly dominant, thus not guaranteeing the clinical relevancy of each mask produced. We introduce SeqSAM, a sequential, RNN-inspired approach to generating multiple masks, which uses a bipartite matching loss for ensuring the clinical relevancy of each mask, and can produce an arbitrary number of masks. We show notable improvements in quality of each mask produced across two publicly available datasets. Our code is available at https://github.com/BenjaminTowle/SeqSAM.
2503.09803
Enes \"Ozeren
Enes \"Ozeren and Arka Bhowmick
Evaluating the Impact of Synthetic Data on Object Detection Tasks in Autonomous Driving
7 pages, 4 figures, 3 tables
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
The increasing applications of autonomous driving systems necessitates large-scale, high-quality datasets to ensure robust performance across diverse scenarios. Synthetic data has emerged as a viable solution to augment real-world datasets due to its cost-effectiveness, availability of precise ground-truth labels, and the ability to model specific edge cases. However, synthetic data may introduce distributional differences and biases that could impact model performance in real-world settings. To evaluate the utility and limitations of synthetic data, we conducted controlled experiments using multiple real-world datasets and a synthetic dataset generated by BIT Technology Solutions GmbH. Our study spans two sensor modalities, camera and LiDAR, and investigates both 2D and 3D object detection tasks. We compare models trained on real, synthetic, and mixed datasets, analyzing their robustness and generalization capabilities. Our findings demonstrate that the use of a combination of real and synthetic data improves the robustness and generalization of object detection models, underscoring the potential of synthetic data in advancing autonomous driving technologies.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 20:13:33 GMT" } ]
2025-03-14T00:00:00
[ [ "Özeren", "Enes", "" ], [ "Bhowmick", "Arka", "" ] ]
TITLE: Evaluating the Impact of Synthetic Data on Object Detection Tasks in Autonomous Driving ABSTRACT: The increasing applications of autonomous driving systems necessitates large-scale, high-quality datasets to ensure robust performance across diverse scenarios. Synthetic data has emerged as a viable solution to augment real-world datasets due to its cost-effectiveness, availability of precise ground-truth labels, and the ability to model specific edge cases. However, synthetic data may introduce distributional differences and biases that could impact model performance in real-world settings. To evaluate the utility and limitations of synthetic data, we conducted controlled experiments using multiple real-world datasets and a synthetic dataset generated by BIT Technology Solutions GmbH. Our study spans two sensor modalities, camera and LiDAR, and investigates both 2D and 3D object detection tasks. We compare models trained on real, synthetic, and mixed datasets, analyzing their robustness and generalization capabilities. Our findings demonstrate that the use of a combination of real and synthetic data improves the robustness and generalization of object detection models, underscoring the potential of synthetic data in advancing autonomous driving technologies.
2503.09807
Qingwu Liu
Qingwu Liu, Nicolas Saunier and Guillaume-Alexandre Bilodeau
How good are deep learning methods for automated road safety analysis using video data? An experimental study
This paper is accepted by TRB Annual Meeting 2024
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
Image-based multi-object detection (MOD) and multi-object tracking (MOT) are advancing at a fast pace. A variety of 2D and 3D MOD and MOT methods have been developed for monocular and stereo cameras. Road safety analysis can benefit from those advancements. As crashes are rare events, surrogate measures of safety (SMoS) have been developed for safety analyses. (Semi-)Automated safety analysis methods extract road user trajectories to compute safety indicators, for example, Time-to-Collision (TTC) and Post-encroachment Time (PET). Inspired by the success of deep learning in MOD and MOT, we investigate three MOT methods, including one based on a stereo-camera, using the annotated KITTI traffic video dataset. Two post-processing steps, IDsplit and SS, are developed to improve the tracking results and investigate the factors influencing the TTC. The experimental results show that, despite some advantages in terms of the numbers of interactions or similarity to the TTC distributions, all the tested methods systematically over-estimate the number of interactions and under-estimate the TTC: they report more interactions and more severe interactions, making the road user interactions appear less safe than they are. Further efforts will be directed towards testing more methods and more data, in particular from roadside sensors, to verify the results and improve the performance.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 20:17:50 GMT" } ]
2025-03-14T00:00:00
[ [ "Liu", "Qingwu", "" ], [ "Saunier", "Nicolas", "" ], [ "Bilodeau", "Guillaume-Alexandre", "" ] ]
TITLE: How good are deep learning methods for automated road safety analysis using video data? An experimental study ABSTRACT: Image-based multi-object detection (MOD) and multi-object tracking (MOT) are advancing at a fast pace. A variety of 2D and 3D MOD and MOT methods have been developed for monocular and stereo cameras. Road safety analysis can benefit from those advancements. As crashes are rare events, surrogate measures of safety (SMoS) have been developed for safety analyses. (Semi-)Automated safety analysis methods extract road user trajectories to compute safety indicators, for example, Time-to-Collision (TTC) and Post-encroachment Time (PET). Inspired by the success of deep learning in MOD and MOT, we investigate three MOT methods, including one based on a stereo-camera, using the annotated KITTI traffic video dataset. Two post-processing steps, IDsplit and SS, are developed to improve the tracking results and investigate the factors influencing the TTC. The experimental results show that, despite some advantages in terms of the numbers of interactions or similarity to the TTC distributions, all the tested methods systematically over-estimate the number of interactions and under-estimate the TTC: they report more interactions and more severe interactions, making the road user interactions appear less safe than they are. Further efforts will be directed towards testing more methods and more data, in particular from roadside sensors, to verify the results and improve the performance.
2503.09808
Chenjun Li
Chenjun Li, Laurin Lux, Alexander H. Berger, Martin J. Menten, Mert R. Sabuncu, Johannes C. Paetzold
Fine-tuning Vision Language Models with Graph-based Knowledge for Explainable Medical Image Analysis
11 pages, 3 figures
null
null
null
cs.CV cs.AI
http://creativecommons.org/licenses/by/4.0/
Accurate staging of Diabetic Retinopathy (DR) is essential for guiding timely interventions and preventing vision loss. However, current staging models are hardly interpretable, and most public datasets contain no clinical reasoning or interpretation beyond image-level labels. In this paper, we present a novel method that integrates graph representation learning with vision-language models (VLMs) to deliver explainable DR diagnosis. Our approach leverages optical coherence tomography angiography (OCTA) images by constructing biologically informed graphs that encode key retinal vascular features such as vessel morphology and spatial connectivity. A graph neural network (GNN) then performs DR staging while integrated gradients highlight critical nodes and edges and their individual features that drive the classification decisions. We collect this graph-based knowledge which attributes the model's prediction to physiological structures and their characteristics. We then transform it into textual descriptions for VLMs. We perform instruction-tuning with these textual descriptions and the corresponding image to train a student VLM. This final agent can classify the disease and explain its decision in a human interpretable way solely based on a single image input. Experimental evaluations on both proprietary and public datasets demonstrate that our method not only improves classification accuracy but also offers more clinically interpretable results. An expert study further demonstrates that our method provides more accurate diagnostic explanations and paves the way for precise localization of pathologies in OCTA images.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 20:19:07 GMT" } ]
2025-03-14T00:00:00
[ [ "Li", "Chenjun", "" ], [ "Lux", "Laurin", "" ], [ "Berger", "Alexander H.", "" ], [ "Menten", "Martin J.", "" ], [ "Sabuncu", "Mert R.", "" ], [ "Paetzold", "Johannes C.", "" ] ]
TITLE: Fine-tuning Vision Language Models with Graph-based Knowledge for Explainable Medical Image Analysis ABSTRACT: Accurate staging of Diabetic Retinopathy (DR) is essential for guiding timely interventions and preventing vision loss. However, current staging models are hardly interpretable, and most public datasets contain no clinical reasoning or interpretation beyond image-level labels. In this paper, we present a novel method that integrates graph representation learning with vision-language models (VLMs) to deliver explainable DR diagnosis. Our approach leverages optical coherence tomography angiography (OCTA) images by constructing biologically informed graphs that encode key retinal vascular features such as vessel morphology and spatial connectivity. A graph neural network (GNN) then performs DR staging while integrated gradients highlight critical nodes and edges and their individual features that drive the classification decisions. We collect this graph-based knowledge which attributes the model's prediction to physiological structures and their characteristics. We then transform it into textual descriptions for VLMs. We perform instruction-tuning with these textual descriptions and the corresponding image to train a student VLM. This final agent can classify the disease and explain its decision in a human interpretable way solely based on a single image input. Experimental evaluations on both proprietary and public datasets demonstrate that our method not only improves classification accuracy but also offers more clinically interpretable results. An expert study further demonstrates that our method provides more accurate diagnostic explanations and paves the way for precise localization of pathologies in OCTA images.
2503.09811
Agata Fronczak
Maciej J. Mrowinski, Aleksandra Buczek, Agata Fronczak
Exploring the dynamics of self-citations and their role in shaping scientific impact
10 pages, 6 figures
null
null
null
cs.DL cs.SI physics.soc-ph
http://creativecommons.org/licenses/by/4.0/
Understanding the mechanisms driving the distribution of scientific citations is a key challenge in assessing the scientific impact of authors. We investigate the influence of the preferential attachment rule (PAR) in this process by analyzing individual citation events from the DBLP dataset, enabling us to estimate the probability of citations being assigned preferentially. Our findings reveal that, for the aggregated dataset, PAR dominates the citation distribution process, with approximately 70% of citations adhering to this mechanism. However, analysis at the individual level shows significant variability, with some authors experiencing a greater prevalence of preferential citations, particularly in the context of external citations. In contrast, self-citations exhibit notably different behaviour, with only 20% following PAR. We also demonstrate that the prominence of PAR increases with an author's citability (average citations per paper), suggesting that more citable authors are preferentially cited, while less-cited authors experience more random citation patterns. Furthermore, we show that self-citations may influence bibliometric indexes. Our results emphasise the distinct dynamics of self-citations compared to external citations, raising questions about the mechanisms driving self-citation patterns. These findings provide new insights into citation behaviours and highlight the limitations of existing approaches in capturing the nuances of scientific impact.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 20:20:45 GMT" } ]
2025-03-14T00:00:00
[ [ "Mrowinski", "Maciej J.", "" ], [ "Buczek", "Aleksandra", "" ], [ "Fronczak", "Agata", "" ] ]
TITLE: Exploring the dynamics of self-citations and their role in shaping scientific impact ABSTRACT: Understanding the mechanisms driving the distribution of scientific citations is a key challenge in assessing the scientific impact of authors. We investigate the influence of the preferential attachment rule (PAR) in this process by analyzing individual citation events from the DBLP dataset, enabling us to estimate the probability of citations being assigned preferentially. Our findings reveal that, for the aggregated dataset, PAR dominates the citation distribution process, with approximately 70% of citations adhering to this mechanism. However, analysis at the individual level shows significant variability, with some authors experiencing a greater prevalence of preferential citations, particularly in the context of external citations. In contrast, self-citations exhibit notably different behaviour, with only 20% following PAR. We also demonstrate that the prominence of PAR increases with an author's citability (average citations per paper), suggesting that more citable authors are preferentially cited, while less-cited authors experience more random citation patterns. Furthermore, we show that self-citations may influence bibliometric indexes. Our results emphasise the distinct dynamics of self-citations compared to external citations, raising questions about the mechanisms driving self-citation patterns. These findings provide new insights into citation behaviours and highlight the limitations of existing approaches in capturing the nuances of scientific impact.
2503.09819
Yuwei Zhang
Yuwei Zhang, Jayanth Srinivasa, Gaowen Liu, Jingbo Shang
Attention Reveals More Than Tokens: Training-Free Long-Context Reasoning with Attention-guided Retrieval
Work in progress
null
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
Large Language Models (LLMs) often exhibit substantially shorter effective context lengths than their claimed capacities, especially when handling complex reasoning tasks that require integrating information from multiple parts of a long context and performing multi-step reasoning. Although Chain-of-Thought (CoT) prompting has shown promise in reducing task complexity, our empirical analysis reveals that it does not fully resolve this limitation. Through controlled experiments, we identify poor recall of implicit facts as the primary cause of failure, which significantly hampers reasoning performance. Interestingly, we observe that the internal attention weights from the generated CoT tokens can effectively ground implicit facts, even when these facts are not explicitly recalled. Building on this insight, we propose a novel training-free algorithm, Attrieval, which leverages attention weights to retrieve relevant facts from the long context and incorporates them into the reasoning process. Additionally, we find that selecting context tokens from CoT tokens further improves performance. Our results demonstrate that Attrieval enhances long-context reasoning capability notably on both synthetic and real-world QA datasets with various models.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 20:34:14 GMT" } ]
2025-03-14T00:00:00
[ [ "Zhang", "Yuwei", "" ], [ "Srinivasa", "Jayanth", "" ], [ "Liu", "Gaowen", "" ], [ "Shang", "Jingbo", "" ] ]
TITLE: Attention Reveals More Than Tokens: Training-Free Long-Context Reasoning with Attention-guided Retrieval ABSTRACT: Large Language Models (LLMs) often exhibit substantially shorter effective context lengths than their claimed capacities, especially when handling complex reasoning tasks that require integrating information from multiple parts of a long context and performing multi-step reasoning. Although Chain-of-Thought (CoT) prompting has shown promise in reducing task complexity, our empirical analysis reveals that it does not fully resolve this limitation. Through controlled experiments, we identify poor recall of implicit facts as the primary cause of failure, which significantly hampers reasoning performance. Interestingly, we observe that the internal attention weights from the generated CoT tokens can effectively ground implicit facts, even when these facts are not explicitly recalled. Building on this insight, we propose a novel training-free algorithm, Attrieval, which leverages attention weights to retrieve relevant facts from the long context and incorporates them into the reasoning process. Additionally, we find that selecting context tokens from CoT tokens further improves performance. Our results demonstrate that Attrieval enhances long-context reasoning capability notably on both synthetic and real-world QA datasets with various models.
2503.09820
Mohamed Elnoor
Mohamed Elnoor, Kasun Weerakoon, Gershom Seneviratne, Jing Liang, Vignesh Rajagopal and Dinesh Manocha
Vi-LAD: Vision-Language Attention Distillation for Socially-Aware Robot Navigation in Dynamic Environments
null
null
null
null
cs.RO cs.AI
http://creativecommons.org/licenses/by/4.0/
We introduce Vision-Language Attention Distillation (Vi-LAD), a novel approach for distilling socially compliant navigation knowledge from a large Vision-Language Model (VLM) into a lightweight transformer model for real-time robotic navigation. Unlike traditional methods that rely on expert demonstrations or human-annotated datasets, Vi-LAD performs knowledge distillation and fine-tuning at the intermediate layer representation level (i.e., attention maps) by leveraging the backbone of a pre-trained vision-action model. These attention maps highlight key navigational regions in a given scene, which serve as implicit guidance for socially aware motion planning. Vi-LAD fine-tunes a transformer-based model using intermediate attention maps extracted from the pre-trained vision-action model, combined with attention-like semantic maps constructed from a large VLM. To achieve this, we introduce a novel attention-level distillation loss that fuses knowledge from both sources, generating augmented attention maps with enhanced social awareness. These refined attention maps are then utilized as a traversability costmap within a socially aware model predictive controller (MPC) for navigation. We validate our approach through real-world experiments on a Husky wheeled robot, demonstrating significant improvements over state-of-the-art (SOTA) navigation methods. Our results show up to 14.2% - 50% improvement in success rate, which highlights the effectiveness of Vi-LAD in enabling socially compliant and efficient robot navigation.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 20:38:23 GMT" } ]
2025-03-14T00:00:00
[ [ "Elnoor", "Mohamed", "" ], [ "Weerakoon", "Kasun", "" ], [ "Seneviratne", "Gershom", "" ], [ "Liang", "Jing", "" ], [ "Rajagopal", "Vignesh", "" ], [ "Manocha", "Dinesh", "" ] ]
TITLE: Vi-LAD: Vision-Language Attention Distillation for Socially-Aware Robot Navigation in Dynamic Environments ABSTRACT: We introduce Vision-Language Attention Distillation (Vi-LAD), a novel approach for distilling socially compliant navigation knowledge from a large Vision-Language Model (VLM) into a lightweight transformer model for real-time robotic navigation. Unlike traditional methods that rely on expert demonstrations or human-annotated datasets, Vi-LAD performs knowledge distillation and fine-tuning at the intermediate layer representation level (i.e., attention maps) by leveraging the backbone of a pre-trained vision-action model. These attention maps highlight key navigational regions in a given scene, which serve as implicit guidance for socially aware motion planning. Vi-LAD fine-tunes a transformer-based model using intermediate attention maps extracted from the pre-trained vision-action model, combined with attention-like semantic maps constructed from a large VLM. To achieve this, we introduce a novel attention-level distillation loss that fuses knowledge from both sources, generating augmented attention maps with enhanced social awareness. These refined attention maps are then utilized as a traversability costmap within a socially aware model predictive controller (MPC) for navigation. We validate our approach through real-world experiments on a Husky wheeled robot, demonstrating significant improvements over state-of-the-art (SOTA) navigation methods. Our results show up to 14.2% - 50% improvement in success rate, which highlights the effectiveness of Vi-LAD in enabling socially compliant and efficient robot navigation.
2503.09826
Wenyi Lian
Wenyi Lian, Joakim Lindblad, Patrick Micke, Nata\v{s}a Sladoje
Isolated Channel Vision Transformers: From Single-Channel Pretraining to Multi-Channel Finetuning
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Vision Transformers (ViTs) have achieved remarkable success in standard RGB image processing tasks. However, applying ViTs to multi-channel imaging (MCI) data, e.g., for medical and remote sensing applications, remains a challenge. In particular, MCI data often consist of layers acquired from different modalities. Directly training ViTs on such data can obscure complementary information and impair the performance. In this paper, we introduce a simple yet effective pretraining framework for large-scale MCI datasets. Our method, named Isolated Channel ViT (IC-ViT), patchifies image channels individually and thereby enables pretraining for multimodal multi-channel tasks. We show that this channel-wise patchifying is a key technique for MCI processing. More importantly, one can pretrain the IC-ViT on single channels and finetune it on downstream multi-channel datasets. This pretraining framework captures dependencies between patches as well as channels and produces robust feature representation. Experiments on various tasks and benchmarks, including JUMP-CP and CHAMMI for cell microscopy imaging, and So2Sat-LCZ42 for satellite imaging, show that the proposed IC-ViT delivers 4-14 percentage points of performance improvement over existing channel-adaptive approaches. Further, its efficient training makes it a suitable candidate for large-scale pretraining of foundation models on heterogeneous data.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 20:45:02 GMT" } ]
2025-03-14T00:00:00
[ [ "Lian", "Wenyi", "" ], [ "Lindblad", "Joakim", "" ], [ "Micke", "Patrick", "" ], [ "Sladoje", "Nataša", "" ] ]
TITLE: Isolated Channel Vision Transformers: From Single-Channel Pretraining to Multi-Channel Finetuning ABSTRACT: Vision Transformers (ViTs) have achieved remarkable success in standard RGB image processing tasks. However, applying ViTs to multi-channel imaging (MCI) data, e.g., for medical and remote sensing applications, remains a challenge. In particular, MCI data often consist of layers acquired from different modalities. Directly training ViTs on such data can obscure complementary information and impair the performance. In this paper, we introduce a simple yet effective pretraining framework for large-scale MCI datasets. Our method, named Isolated Channel ViT (IC-ViT), patchifies image channels individually and thereby enables pretraining for multimodal multi-channel tasks. We show that this channel-wise patchifying is a key technique for MCI processing. More importantly, one can pretrain the IC-ViT on single channels and finetune it on downstream multi-channel datasets. This pretraining framework captures dependencies between patches as well as channels and produces robust feature representation. Experiments on various tasks and benchmarks, including JUMP-CP and CHAMMI for cell microscopy imaging, and So2Sat-LCZ42 for satellite imaging, show that the proposed IC-ViT delivers 4-14 percentage points of performance improvement over existing channel-adaptive approaches. Further, its efficient training makes it a suitable candidate for large-scale pretraining of foundation models on heterogeneous data.
2503.09850
Ali Eslamian
Ali Eslamian, Qiang Cheng
TabNSA: Native Sparse Attention for Efficient Tabular Data Learning
5 pages, 4 tables
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Tabular data poses unique challenges for deep learning due to its heterogeneous features and lack of inherent spatial structure. This paper introduces TabNSA, a novel deep learning architecture leveraging Native Sparse Attention (NSA) specifically for efficient tabular data processing. TabNSA incorporates a dynamic hierarchical sparse strategy, combining coarse-grained feature compression with fine-grained feature selection to preserve both global context awareness and local precision. By dynamically focusing on relevant subsets of features, TabNSA effectively captures intricate feature interactions. Extensive experiments demonstrate that TabNSA consistently outperforms existing methods, including both deep learning architectures and ensemble decision trees, achieving state-of-the-art performance across various benchmark datasets.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 21:13:41 GMT" } ]
2025-03-14T00:00:00
[ [ "Eslamian", "Ali", "" ], [ "Cheng", "Qiang", "" ] ]
TITLE: TabNSA: Native Sparse Attention for Efficient Tabular Data Learning ABSTRACT: Tabular data poses unique challenges for deep learning due to its heterogeneous features and lack of inherent spatial structure. This paper introduces TabNSA, a novel deep learning architecture leveraging Native Sparse Attention (NSA) specifically for efficient tabular data processing. TabNSA incorporates a dynamic hierarchical sparse strategy, combining coarse-grained feature compression with fine-grained feature selection to preserve both global context awareness and local precision. By dynamically focusing on relevant subsets of features, TabNSA effectively captures intricate feature interactions. Extensive experiments demonstrate that TabNSA consistently outperforms existing methods, including both deep learning architectures and ensemble decision trees, achieving state-of-the-art performance across various benchmark datasets.
2503.09852
An Yang
An Yang, Chenyu Liu, Pengcheng Xia, Jun Du
StyleSpeaker: Audio-Enhanced Fine-Grained Style Modeling for Speech-Driven 3D Facial Animation
null
null
null
null
cs.MM
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Speech-driven 3D facial animation is challenging due to the diversity in speaking styles and the limited availability of 3D audio-visual data. Speech predominantly dictates the coarse motion trends of the lip region, while specific styles determine the details of lip motion and the overall facial expressions. Prior works lack fine-grained learning in style modeling and do not adequately consider style biases across varying speech conditions, which reduce the accuracy of style modeling and hamper the adaptation capability to unseen speakers. To address this, we propose a novel framework, StyleSpeaker, which explicitly extracts speaking styles based on speaker characteristics while accounting for style biases caused by different speeches. Specifically, we utilize a style encoder to capture speakers' styles from facial motions and enhance them according to motion preferences elicited by varying speech conditions. The enhanced styles are then integrated into the coarse motion features via a style infusion module, which employs a set of style primitives to learn fine-grained style representation. Throughout training, we maintain this set of style primitives to comprehensively model the entire style space. Hence, StyleSpeaker possesses robust style modeling capability for seen speakers and can rapidly adapt to unseen speakers without fine-tuning. Additionally, we design a trend loss and a local contrastive loss to improve the synchronization between synthesized motions and speeches. Extensive qualitative and quantitative experiments on three public datasets demonstrate that our method outperforms existing state-of-the-art approaches.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 21:18:20 GMT" } ]
2025-03-14T00:00:00
[ [ "Yang", "An", "" ], [ "Liu", "Chenyu", "" ], [ "Xia", "Pengcheng", "" ], [ "Du", "Jun", "" ] ]
TITLE: StyleSpeaker: Audio-Enhanced Fine-Grained Style Modeling for Speech-Driven 3D Facial Animation ABSTRACT: Speech-driven 3D facial animation is challenging due to the diversity in speaking styles and the limited availability of 3D audio-visual data. Speech predominantly dictates the coarse motion trends of the lip region, while specific styles determine the details of lip motion and the overall facial expressions. Prior works lack fine-grained learning in style modeling and do not adequately consider style biases across varying speech conditions, which reduce the accuracy of style modeling and hamper the adaptation capability to unseen speakers. To address this, we propose a novel framework, StyleSpeaker, which explicitly extracts speaking styles based on speaker characteristics while accounting for style biases caused by different speeches. Specifically, we utilize a style encoder to capture speakers' styles from facial motions and enhance them according to motion preferences elicited by varying speech conditions. The enhanced styles are then integrated into the coarse motion features via a style infusion module, which employs a set of style primitives to learn fine-grained style representation. Throughout training, we maintain this set of style primitives to comprehensively model the entire style space. Hence, StyleSpeaker possesses robust style modeling capability for seen speakers and can rapidly adapt to unseen speakers without fine-tuning. Additionally, we design a trend loss and a local contrastive loss to improve the synchronization between synthesized motions and speeches. Extensive qualitative and quantitative experiments on three public datasets demonstrate that our method outperforms existing state-of-the-art approaches.
2503.09860
Nahid Ul Islam
Nahid Ul Islam, DongAo Ma, Jiaxuan Pang, Shivasakthi Senthil Velan, Michael Gotway, Jianming Liang
Foundation X: Integrating Classification, Localization, and Segmentation through Lock-Release Pretraining Strategy for Chest X-ray Analysis
Accepted by WACV 2025
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Developing robust and versatile deep-learning models is essential for enhancing diagnostic accuracy and guiding clinical interventions in medical imaging, but it requires a large amount of annotated data. The advancement of deep learning has facilitated the creation of numerous medical datasets with diverse expert-level annotations. Aggregating these datasets can maximize data utilization and address the inadequacy of labeled data. However, the heterogeneity of expert-level annotations across tasks such as classification, localization, and segmentation presents a significant challenge for learning from these datasets. To this end, we introduce nFoundation X, an end-to-end framework that utilizes diverse expert-level annotations from numerous public datasets to train a foundation model capable of multiple tasks including classification, localization, and segmentation. To address the challenges of annotation and task heterogeneity, we propose a Lock-Release pretraining strategy to enhance the cyclic learning from multiple datasets, combined with the student-teacher learning paradigm, ensuring the model retains general knowledge for all tasks while preventing overfitting to any single task. To demonstrate the effectiveness of Foundation X, we trained a model using 11 chest X-ray datasets, covering annotations for classification, localization, and segmentation tasks. Our experimental results show that Foundation X achieves notable performance gains through extensive annotation utilization, excels in cross-dataset and cross-task learning, and further enhances performance in organ localization and segmentation tasks. All code and pretrained models are publicly accessible at https://github.com/jlianglab/Foundation_X.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 21:45:13 GMT" } ]
2025-03-14T00:00:00
[ [ "Islam", "Nahid Ul", "" ], [ "Ma", "DongAo", "" ], [ "Pang", "Jiaxuan", "" ], [ "Velan", "Shivasakthi Senthil", "" ], [ "Gotway", "Michael", "" ], [ "Liang", "Jianming", "" ] ]
TITLE: Foundation X: Integrating Classification, Localization, and Segmentation through Lock-Release Pretraining Strategy for Chest X-ray Analysis ABSTRACT: Developing robust and versatile deep-learning models is essential for enhancing diagnostic accuracy and guiding clinical interventions in medical imaging, but it requires a large amount of annotated data. The advancement of deep learning has facilitated the creation of numerous medical datasets with diverse expert-level annotations. Aggregating these datasets can maximize data utilization and address the inadequacy of labeled data. However, the heterogeneity of expert-level annotations across tasks such as classification, localization, and segmentation presents a significant challenge for learning from these datasets. To this end, we introduce nFoundation X, an end-to-end framework that utilizes diverse expert-level annotations from numerous public datasets to train a foundation model capable of multiple tasks including classification, localization, and segmentation. To address the challenges of annotation and task heterogeneity, we propose a Lock-Release pretraining strategy to enhance the cyclic learning from multiple datasets, combined with the student-teacher learning paradigm, ensuring the model retains general knowledge for all tasks while preventing overfitting to any single task. To demonstrate the effectiveness of Foundation X, we trained a model using 11 chest X-ray datasets, covering annotations for classification, localization, and segmentation tasks. Our experimental results show that Foundation X achieves notable performance gains through extensive annotation utilization, excels in cross-dataset and cross-task learning, and further enhances performance in organ localization and segmentation tasks. All code and pretrained models are publicly accessible at https://github.com/jlianglab/Foundation_X.
2503.09868
Phil Travis
Phil Travis, Jacob Bortnik, Troy Carter
Machine-learned trends in mirror configurations in the Large Plasma Device
16 pages, 11 figures
null
null
null
physics.plasm-ph
http://creativecommons.org/licenses/by-nc-sa/4.0/
This study demonstrates the efficacy of ML-based trend inference using data from the Large Plasma Device (LAPD). The LAPD is a flexible basic plasma science device with a high discharge repetition rate (0.25-1 Hz) and reproducible plasmas capable of collecting high-spatial-resolution probe measurements. A diverse dataset is collected through random sampling of LAPD operational parameters, including the magnetic field strength and profile, fueling settings, and the discharge voltage. NN ensembles with uncertainty quantification are trained to predict time-averaged ion saturation current ($I_\text{sat}$ -- proportional to density and the square root of electron temperature) at any position within the dataset domain. Model-inferred trends, such as the effects of introducing mirrors or changing the discharge voltage, are consistent with current understanding. In addition, axial variation is optimized via comprehensive search over $I_\text{sat}$ predictions. Experimental validation of these optimized machine parameters demonstrate qualitative agreement, with quantitative differences attributable to Langmuir probe variation and cathode conditions. This investigation demonstrates, using ML techniques, a new way of extracting insight from experiments and novel optimization of plasmas. The code and data used in this study are made freely available.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 22:00:48 GMT" } ]
2025-03-14T00:00:00
[ [ "Travis", "Phil", "" ], [ "Bortnik", "Jacob", "" ], [ "Carter", "Troy", "" ] ]
TITLE: Machine-learned trends in mirror configurations in the Large Plasma Device ABSTRACT: This study demonstrates the efficacy of ML-based trend inference using data from the Large Plasma Device (LAPD). The LAPD is a flexible basic plasma science device with a high discharge repetition rate (0.25-1 Hz) and reproducible plasmas capable of collecting high-spatial-resolution probe measurements. A diverse dataset is collected through random sampling of LAPD operational parameters, including the magnetic field strength and profile, fueling settings, and the discharge voltage. NN ensembles with uncertainty quantification are trained to predict time-averaged ion saturation current ($I_\text{sat}$ -- proportional to density and the square root of electron temperature) at any position within the dataset domain. Model-inferred trends, such as the effects of introducing mirrors or changing the discharge voltage, are consistent with current understanding. In addition, axial variation is optimized via comprehensive search over $I_\text{sat}$ predictions. Experimental validation of these optimized machine parameters demonstrate qualitative agreement, with quantitative differences attributable to Langmuir probe variation and cathode conditions. This investigation demonstrates, using ML techniques, a new way of extracting insight from experiments and novel optimization of plasmas. The code and data used in this study are made freely available.
2503.09873
Shoaib Meraj Sami
Shoaib Meraj Sami, Md Mahedi Hasan, Nasser M. Nasrabadi, Raghuveer Rao
FDCT: Frequency-Aware Decomposition and Cross-Modal Token-Alignment for Multi-Sensor Target Classification
12 pages Accepted in the IEEE Transactions on Aerospace and Electronic Systems
null
10.1109/TAES.2025.3550474
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
In automatic target recognition (ATR) systems, sensors may fail to capture discriminative, fine-grained detail features due to environmental conditions, noise created by CMOS chips, occlusion, parallaxes, and sensor misalignment. Therefore, multi-sensor image fusion is an effective choice to overcome these constraints. However, multi-modal image sensors are heterogeneous and have domain and granularity gaps. In addition, the multi-sensor images can be misaligned due to intricate background clutters, fluctuating illumination conditions, and uncontrolled sensor settings. In this paper, to overcome these issues, we decompose, align, and fuse multiple image sensor data for target classification. We extract the domain-specific and domain-invariant features from each sensor data. We propose to develop a shared unified discrete token (UDT) space between sensors to reduce the domain and granularity gaps. Additionally, we develop an alignment module to overcome the misalignment between multi-sensors and emphasize the discriminative representation of the UDT space. In the alignment module, we introduce sparsity constraints to provide a better cross-modal representation of the UDT space and robustness against various sensor settings. We achieve superior classification performance compared to single-modality classifiers and several state-of-the-art multi-modal fusion algorithms on four multi-sensor ATR datasets.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 22:12:35 GMT" } ]
2025-03-14T00:00:00
[ [ "Sami", "Shoaib Meraj", "" ], [ "Hasan", "Md Mahedi", "" ], [ "Nasrabadi", "Nasser M.", "" ], [ "Rao", "Raghuveer", "" ] ]
TITLE: FDCT: Frequency-Aware Decomposition and Cross-Modal Token-Alignment for Multi-Sensor Target Classification ABSTRACT: In automatic target recognition (ATR) systems, sensors may fail to capture discriminative, fine-grained detail features due to environmental conditions, noise created by CMOS chips, occlusion, parallaxes, and sensor misalignment. Therefore, multi-sensor image fusion is an effective choice to overcome these constraints. However, multi-modal image sensors are heterogeneous and have domain and granularity gaps. In addition, the multi-sensor images can be misaligned due to intricate background clutters, fluctuating illumination conditions, and uncontrolled sensor settings. In this paper, to overcome these issues, we decompose, align, and fuse multiple image sensor data for target classification. We extract the domain-specific and domain-invariant features from each sensor data. We propose to develop a shared unified discrete token (UDT) space between sensors to reduce the domain and granularity gaps. Additionally, we develop an alignment module to overcome the misalignment between multi-sensors and emphasize the discriminative representation of the UDT space. In the alignment module, we introduce sparsity constraints to provide a better cross-modal representation of the UDT space and robustness against various sensor settings. We achieve superior classification performance compared to single-modality classifiers and several state-of-the-art multi-modal fusion algorithms on four multi-sensor ATR datasets.
2503.09896
Ping Chen Dr.
Ping Chen, David Hinote, Guoqing Chen
A Rule Based Solution to Co-reference Resolution in Clinical Text
null
null
null
null
cs.CL cs.AI
http://creativecommons.org/licenses/by/4.0/
Objective: The aim of this study was to build an effective co-reference resolution system tailored for the biomedical domain. Materials and Methods: Experiment materials used in this study is provided by the 2011 i2b2 Natural Language Processing Challenge. The 2011 i2b2 challenge involves coreference resolution in medical documents. Concept mentions have been annotated in clinical texts, and the mentions that co-refer in each document are to be linked by coreference chains. Normally, there are two ways of constructing a system to automatically discover co-referent links. One is to manually build rules for co-reference resolution, and the other category of approaches is to use machine learning systems to learn automatically from training datasets and then perform the resolution task on testing datasets. Results: Experiments show the existing co-reference resolution systems are able to find some of the co-referent links, and our rule based system performs well finding the majority of the co-referent links. Our system achieved 89.6% overall performance on multiple medical datasets. Conclusion: The experiment results show that manually crafted rules based on observation of training data is a valid way to accomplish high performance in this coreference resolution task for the critical biomedical domain.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 23:29:08 GMT" } ]
2025-03-14T00:00:00
[ [ "Chen", "Ping", "" ], [ "Hinote", "David", "" ], [ "Chen", "Guoqing", "" ] ]
TITLE: A Rule Based Solution to Co-reference Resolution in Clinical Text ABSTRACT: Objective: The aim of this study was to build an effective co-reference resolution system tailored for the biomedical domain. Materials and Methods: Experiment materials used in this study is provided by the 2011 i2b2 Natural Language Processing Challenge. The 2011 i2b2 challenge involves coreference resolution in medical documents. Concept mentions have been annotated in clinical texts, and the mentions that co-refer in each document are to be linked by coreference chains. Normally, there are two ways of constructing a system to automatically discover co-referent links. One is to manually build rules for co-reference resolution, and the other category of approaches is to use machine learning systems to learn automatically from training datasets and then perform the resolution task on testing datasets. Results: Experiments show the existing co-reference resolution systems are able to find some of the co-referent links, and our rule based system performs well finding the majority of the co-referent links. Our system achieved 89.6% overall performance on multiple medical datasets. Conclusion: The experiment results show that manually crafted rules based on observation of training data is a valid way to accomplish high performance in this coreference resolution task for the critical biomedical domain.
2503.09902
Zahra Abbasiantaeb
Zahra Abbasiantaeb, Simon Lupart, Leif Azzopardi, Jeffery Dalton, Mohammad Aliannejadi
Conversational Gold: Evaluating Personalized Conversational Search System using Gold Nuggets
null
null
null
null
cs.IR
http://creativecommons.org/licenses/by/4.0/
The rise of personalized conversational search systems has been driven by advancements in Large Language Models (LLMs), enabling these systems to retrieve and generate answers for complex information needs. However, the automatic evaluation of responses generated by Retrieval Augmented Generation (RAG) systems remains an understudied challenge. In this paper, we introduce a new resource for assessing the retrieval effectiveness and relevance of response generated by RAG systems, using a nugget-based evaluation framework. Built upon the foundation of TREC iKAT 2023, our dataset extends to the TREC iKAT 2024 collection, which includes 17 conversations and 20,575 relevance passage assessments, together with 2,279 extracted gold nuggets, and 62 manually written gold answers from NIST assessors. While maintaining the core structure of its predecessor, this new collection enables a deeper exploration of generation tasks in conversational settings. Key improvements in iKAT 2024 include: (1) ``gold nuggets'' -- concise, essential pieces of information extracted from relevant passages of the collection -- which serve as a foundation for automatic response evaluation; (2) manually written answers to provide a gold standard for response evaluation; (3) unanswerable questions to evaluate model hallucination; (4) expanded user personas, providing richer contextual grounding; and (5) a transition from Personal Text Knowledge Base (PTKB) ranking to PTKB classification and selection. Built on this resource, we provide a framework for long-form answer generation evaluation, involving nuggets extraction and nuggets matching, linked to retrieval. This establishes a solid resource for advancing research in personalized conversational search and long-form answer generation. Our resources are publicly available at https://github.com/irlabamsterdam/CONE-RAG.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 23:44:10 GMT" } ]
2025-03-14T00:00:00
[ [ "Abbasiantaeb", "Zahra", "" ], [ "Lupart", "Simon", "" ], [ "Azzopardi", "Leif", "" ], [ "Dalton", "Jeffery", "" ], [ "Aliannejadi", "Mohammad", "" ] ]
TITLE: Conversational Gold: Evaluating Personalized Conversational Search System using Gold Nuggets ABSTRACT: The rise of personalized conversational search systems has been driven by advancements in Large Language Models (LLMs), enabling these systems to retrieve and generate answers for complex information needs. However, the automatic evaluation of responses generated by Retrieval Augmented Generation (RAG) systems remains an understudied challenge. In this paper, we introduce a new resource for assessing the retrieval effectiveness and relevance of response generated by RAG systems, using a nugget-based evaluation framework. Built upon the foundation of TREC iKAT 2023, our dataset extends to the TREC iKAT 2024 collection, which includes 17 conversations and 20,575 relevance passage assessments, together with 2,279 extracted gold nuggets, and 62 manually written gold answers from NIST assessors. While maintaining the core structure of its predecessor, this new collection enables a deeper exploration of generation tasks in conversational settings. Key improvements in iKAT 2024 include: (1) ``gold nuggets'' -- concise, essential pieces of information extracted from relevant passages of the collection -- which serve as a foundation for automatic response evaluation; (2) manually written answers to provide a gold standard for response evaluation; (3) unanswerable questions to evaluate model hallucination; (4) expanded user personas, providing richer contextual grounding; and (5) a transition from Personal Text Knowledge Base (PTKB) ranking to PTKB classification and selection. Built on this resource, we provide a framework for long-form answer generation evaluation, involving nuggets extraction and nuggets matching, linked to retrieval. This establishes a solid resource for advancing research in personalized conversational search and long-form answer generation. Our resources are publicly available at https://github.com/irlabamsterdam/CONE-RAG.
2503.09903
Ti Nguyen
Ti Ti Nguyen, Thanh-Dung Le, Vu Nguyen Ha, Hong-fu Chou, Geoffrey Eappen, Duc-Dung Tran, Hung Nguyen-Kha, Prabhu Thiruvasagam, Luis M. Garces-Socarras, Jorge L. Gonzalez-Rios, Juan C. Merlano-Duncan, Symeon Chatzinotas
A Semantic-Loss Function Modeling Framework With Task-Oriented Machine Learning Perspectives
6 pages, 11 figures
null
null
null
cs.LG math.OC
http://creativecommons.org/licenses/by/4.0/
The integration of machine learning (ML) has significantly enhanced the capabilities of Earth Observation (EO) systems by enabling the extraction of actionable insights from complex datasets. However, the performance of data-driven EO applications is heavily influenced by the data collection and transmission processes, where limited satellite bandwidth and latency constraints can hinder the full transmission of original data to the receivers. To address this issue, adopting the concepts of Semantic Communication (SC) offers a promising solution by prioritizing the transmission of essential data semantics over raw information. Implementing SC for EO systems requires a thorough understanding of the impact of data processing and communication channel conditions on semantic loss at the processing center. This work proposes a novel data-fitting framework to empirically model the semantic loss using real-world EO datasets and domain-specific insights. The framework quantifies two primary types of semantic loss: (1) source coding loss, assessed via a data quality indicator measuring the impact of processing on raw source data, and (2) transmission loss, evaluated by comparing practical transmission performance against the Shannon limit. Semantic losses are estimated by evaluating the accuracy of EO applications using four task-oriented ML models, EfficientViT, MobileViT, ResNet50-DINO, and ResNet8-KD, on lossy image datasets under varying channel conditions and compression ratios. These results underpin a framework for efficient semantic-loss modeling in bandwidth-constrained EO scenarios, enabling more reliable and effective operations.
[ { "version": "v1", "created": "Wed, 12 Mar 2025 23:45:11 GMT" } ]
2025-03-14T00:00:00
[ [ "Nguyen", "Ti Ti", "" ], [ "Le", "Thanh-Dung", "" ], [ "Ha", "Vu Nguyen", "" ], [ "Chou", "Hong-fu", "" ], [ "Eappen", "Geoffrey", "" ], [ "Tran", "Duc-Dung", "" ], [ "Nguyen-Kha", "Hung", "" ], [ "...
TITLE: A Semantic-Loss Function Modeling Framework With Task-Oriented Machine Learning Perspectives ABSTRACT: The integration of machine learning (ML) has significantly enhanced the capabilities of Earth Observation (EO) systems by enabling the extraction of actionable insights from complex datasets. However, the performance of data-driven EO applications is heavily influenced by the data collection and transmission processes, where limited satellite bandwidth and latency constraints can hinder the full transmission of original data to the receivers. To address this issue, adopting the concepts of Semantic Communication (SC) offers a promising solution by prioritizing the transmission of essential data semantics over raw information. Implementing SC for EO systems requires a thorough understanding of the impact of data processing and communication channel conditions on semantic loss at the processing center. This work proposes a novel data-fitting framework to empirically model the semantic loss using real-world EO datasets and domain-specific insights. The framework quantifies two primary types of semantic loss: (1) source coding loss, assessed via a data quality indicator measuring the impact of processing on raw source data, and (2) transmission loss, evaluated by comparing practical transmission performance against the Shannon limit. Semantic losses are estimated by evaluating the accuracy of EO applications using four task-oriented ML models, EfficientViT, MobileViT, ResNet50-DINO, and ResNet8-KD, on lossy image datasets under varying channel conditions and compression ratios. These results underpin a framework for efficient semantic-loss modeling in bandwidth-constrained EO scenarios, enabling more reliable and effective operations.
2503.09905
Allison Andreyev
Allison Andreyev
Quantization for OpenAI's Whisper Models: A Comparative Analysis
7 pages
null
null
null
cs.SD cs.CL cs.LG eess.AS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Automated speech recognition (ASR) models have gained prominence for applications such as captioning, speech translation, and live transcription. This paper studies Whisper and two model variants: one optimized for live speech streaming and another for offline transcription. Notably, these models have been found to generate hallucinated content, reducing transcription reliability. Furthermore, larger model variants exhibit increased latency and pose challenges for deployment on resource-constrained devices. This study analyzes the similarities and differences between three Whisper models, qualitatively examining their distinct capabilities. Next, this study quantifies the impact of model quantization on latency and evaluates its viability for edge deployment. Using the open source LibriSpeech dataset, this paper evaluates the word error rate (WER) along with latency analysis of whispercpp using 3 quantization methods (INT4, INT5, INT8). Results show that quantization reduces latency by 19\% and model size by 45\%, while preserving transcription accuracy. These findings provide insights into the optimal use cases of different Whisper models and edge device deployment possibilities. All code, datasets, and implementation details are available in a public GitHub repository: https://github.com/allisonandreyev/WhisperQuantization.git
[ { "version": "v1", "created": "Wed, 12 Mar 2025 23:50:35 GMT" } ]
2025-03-14T00:00:00
[ [ "Andreyev", "Allison", "" ] ]
TITLE: Quantization for OpenAI's Whisper Models: A Comparative Analysis ABSTRACT: Automated speech recognition (ASR) models have gained prominence for applications such as captioning, speech translation, and live transcription. This paper studies Whisper and two model variants: one optimized for live speech streaming and another for offline transcription. Notably, these models have been found to generate hallucinated content, reducing transcription reliability. Furthermore, larger model variants exhibit increased latency and pose challenges for deployment on resource-constrained devices. This study analyzes the similarities and differences between three Whisper models, qualitatively examining their distinct capabilities. Next, this study quantifies the impact of model quantization on latency and evaluates its viability for edge deployment. Using the open source LibriSpeech dataset, this paper evaluates the word error rate (WER) along with latency analysis of whispercpp using 3 quantization methods (INT4, INT5, INT8). Results show that quantization reduces latency by 19\% and model size by 45\%, while preserving transcription accuracy. These findings provide insights into the optimal use cases of different Whisper models and edge device deployment possibilities. All code, datasets, and implementation details are available in a public GitHub repository: https://github.com/allisonandreyev/WhisperQuantization.git
2503.09911
Kohei Hayashi
Kohei Hayashi, Masanori Koyama, Julian Jorge Andrade Guerreiro
Inter-environmental world modeling for continuous and compositional dynamics
null
null
null
null
cs.LG
http://creativecommons.org/licenses/by/4.0/
Various world model frameworks are being developed today based on autoregressive frameworks that rely on discrete representations of actions and observations, and these frameworks are succeeding in constructing interactive generative models for the target environment of interest. Meanwhile, humans demonstrate remarkable generalization abilities to combine experiences in multiple environments to mentally simulate and learn to control agents in diverse environments. Inspired by this human capability, we introduce World modeling through Lie Action (WLA), an unsupervised framework that learns continuous latent action representations to simulate across environments. WLA learns a control interface with high controllability and predictive ability by simultaneously modeling the dynamics of multiple environments using Lie group theory and object-centric autoencoder. On synthetic benchmark and real-world datasets, we demonstrate that WLA can be trained using only video frames and, with minimal or no action labels, can quickly adapt to new environments with novel action sets.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 00:02:54 GMT" } ]
2025-03-14T00:00:00
[ [ "Hayashi", "Kohei", "" ], [ "Koyama", "Masanori", "" ], [ "Guerreiro", "Julian Jorge Andrade", "" ] ]
TITLE: Inter-environmental world modeling for continuous and compositional dynamics ABSTRACT: Various world model frameworks are being developed today based on autoregressive frameworks that rely on discrete representations of actions and observations, and these frameworks are succeeding in constructing interactive generative models for the target environment of interest. Meanwhile, humans demonstrate remarkable generalization abilities to combine experiences in multiple environments to mentally simulate and learn to control agents in diverse environments. Inspired by this human capability, we introduce World modeling through Lie Action (WLA), an unsupervised framework that learns continuous latent action representations to simulate across environments. WLA learns a control interface with high controllability and predictive ability by simultaneously modeling the dynamics of multiple environments using Lie group theory and object-centric autoencoder. On synthetic benchmark and real-world datasets, we demonstrate that WLA can be trained using only video frames and, with minimal or no action labels, can quickly adapt to new environments with novel action sets.
2503.09929
Weiwei Zhou
Weiwei Zhou, Chenkun Ling, Zefeng Cai
Emotion Recognition with CLIP and Sequential Learning
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
Human emotion recognition plays a crucial role in facilitating seamless interactions between humans and computers. In this paper, we present our innovative methodology for tackling the Valence-Arousal (VA) Estimation Challenge, the Expression Recognition Challenge, and the Action Unit (AU) Detection Challenge, all within the framework of the 8th Workshop and Competition on Affective Behavior Analysis in-the-wild (ABAW). Our approach introduces a novel framework aimed at enhancing continuous emotion recognition. This is achieved by fine-tuning the CLIP model with the aff-wild2 dataset, which provides annotated expression labels. The result is a fine-tuned model that serves as an efficient visual feature extractor, significantly improving its robustness. To further boost the performance of continuous emotion recognition, we incorporate Temporal Convolutional Network (TCN) modules alongside Transformer Encoder modules into our system architecture. The integration of these advanced components allows our model to outperform baseline performance, demonstrating its ability to recognize human emotions with greater accuracy and efficiency.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 01:02:06 GMT" } ]
2025-03-14T00:00:00
[ [ "Zhou", "Weiwei", "" ], [ "Ling", "Chenkun", "" ], [ "Cai", "Zefeng", "" ] ]
TITLE: Emotion Recognition with CLIP and Sequential Learning ABSTRACT: Human emotion recognition plays a crucial role in facilitating seamless interactions between humans and computers. In this paper, we present our innovative methodology for tackling the Valence-Arousal (VA) Estimation Challenge, the Expression Recognition Challenge, and the Action Unit (AU) Detection Challenge, all within the framework of the 8th Workshop and Competition on Affective Behavior Analysis in-the-wild (ABAW). Our approach introduces a novel framework aimed at enhancing continuous emotion recognition. This is achieved by fine-tuning the CLIP model with the aff-wild2 dataset, which provides annotated expression labels. The result is a fine-tuned model that serves as an efficient visual feature extractor, significantly improving its robustness. To further boost the performance of continuous emotion recognition, we incorporate Temporal Convolutional Network (TCN) modules alongside Transformer Encoder modules into our system architecture. The integration of these advanced components allows our model to outperform baseline performance, demonstrating its ability to recognize human emotions with greater accuracy and efficiency.
2503.09938
Dongliang Zhou
Sen Wang, Dongliang Zhou, Liang Xie, Chao Xu, Ye Yan, Erwei Yin
PanoGen++: Domain-Adapted Text-Guided Panoramic Environment Generation for Vision-and-Language Navigation
This paper was accepted by Neural Networks
null
10.1016/j.neunet.2025.107320
null
cs.CV cs.MM cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Vision-and-language navigation (VLN) tasks require agents to navigate three-dimensional environments guided by natural language instructions, offering substantial potential for diverse applications. However, the scarcity of training data impedes progress in this field. This paper introduces PanoGen++, a novel framework that addresses this limitation by generating varied and pertinent panoramic environments for VLN tasks. PanoGen++ incorporates pre-trained diffusion models with domain-specific fine-tuning, employing parameter-efficient techniques such as low-rank adaptation to minimize computational costs. We investigate two settings for environment generation: masked image inpainting and recursive image outpainting. The former maximizes novel environment creation by inpainting masked regions based on textual descriptions, while the latter facilitates agents' learning of spatial relationships within panoramas. Empirical evaluations on room-to-room (R2R), room-for-room (R4R), and cooperative vision-and-dialog navigation (CVDN) datasets reveal significant performance enhancements: a 2.44% increase in success rate on the R2R test leaderboard, a 0.63% improvement on the R4R validation unseen set, and a 0.75-meter enhancement in goal progress on the CVDN validation unseen set. PanoGen++ augments the diversity and relevance of training environments, resulting in improved generalization and efficacy in VLN tasks.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 01:16:58 GMT" } ]
2025-03-14T00:00:00
[ [ "Wang", "Sen", "" ], [ "Zhou", "Dongliang", "" ], [ "Xie", "Liang", "" ], [ "Xu", "Chao", "" ], [ "Yan", "Ye", "" ], [ "Yin", "Erwei", "" ] ]
TITLE: PanoGen++: Domain-Adapted Text-Guided Panoramic Environment Generation for Vision-and-Language Navigation ABSTRACT: Vision-and-language navigation (VLN) tasks require agents to navigate three-dimensional environments guided by natural language instructions, offering substantial potential for diverse applications. However, the scarcity of training data impedes progress in this field. This paper introduces PanoGen++, a novel framework that addresses this limitation by generating varied and pertinent panoramic environments for VLN tasks. PanoGen++ incorporates pre-trained diffusion models with domain-specific fine-tuning, employing parameter-efficient techniques such as low-rank adaptation to minimize computational costs. We investigate two settings for environment generation: masked image inpainting and recursive image outpainting. The former maximizes novel environment creation by inpainting masked regions based on textual descriptions, while the latter facilitates agents' learning of spatial relationships within panoramas. Empirical evaluations on room-to-room (R2R), room-for-room (R4R), and cooperative vision-and-dialog navigation (CVDN) datasets reveal significant performance enhancements: a 2.44% increase in success rate on the R2R test leaderboard, a 0.63% improvement on the R4R validation unseen set, and a 0.75-meter enhancement in goal progress on the CVDN validation unseen set. PanoGen++ augments the diversity and relevance of training environments, resulting in improved generalization and efficacy in VLN tasks.
2503.09941
Wenyu Chen
Mu Chen, Wenyu Chen, Mingchuan Yang, Yuan Zhang, Tao Han, Xinchi Li, Yunlong Li, Huaici Zhao
TGP: Two-modal occupancy prediction with 3D Gaussian and sparse points for 3D Environment Awareness
null
null
null
null
cs.CV cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
3D semantic occupancy has rapidly become a research focus in the fields of robotics and autonomous driving environment perception due to its ability to provide more realistic geometric perception and its closer integration with downstream tasks. By performing occupancy prediction of the 3D space in the environment, the ability and robustness of scene understanding can be effectively improved. However, existing occupancy prediction tasks are primarily modeled using voxel or point cloud-based approaches: voxel-based network structures often suffer from the loss of spatial information due to the voxelization process, while point cloud-based methods, although better at retaining spatial location information, face limitations in representing volumetric structural details. To address this issue, we propose a dual-modal prediction method based on 3D Gaussian sets and sparse points, which balances both spatial location and volumetric structural information, achieving higher accuracy in semantic occupancy prediction. Specifically, our method adopts a Transformer-based architecture, taking 3D Gaussian sets, sparse points, and queries as inputs. Through the multi-layer structure of the Transformer, the enhanced queries and 3D Gaussian sets jointly contribute to the semantic occupancy prediction, and an adaptive fusion mechanism integrates the semantic outputs of both modalities to generate the final prediction results. Additionally, to further improve accuracy, we dynamically refine the point cloud at each layer, allowing for more precise location information during occupancy prediction. We conducted experiments on the Occ3DnuScenes dataset, and the experimental results demonstrate superior performance of the proposed method on IoU based metrics.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 01:35:04 GMT" } ]
2025-03-14T00:00:00
[ [ "Chen", "Mu", "" ], [ "Chen", "Wenyu", "" ], [ "Yang", "Mingchuan", "" ], [ "Zhang", "Yuan", "" ], [ "Han", "Tao", "" ], [ "Li", "Xinchi", "" ], [ "Li", "Yunlong", "" ], [ "Zhao", "Huaici", ...
TITLE: TGP: Two-modal occupancy prediction with 3D Gaussian and sparse points for 3D Environment Awareness ABSTRACT: 3D semantic occupancy has rapidly become a research focus in the fields of robotics and autonomous driving environment perception due to its ability to provide more realistic geometric perception and its closer integration with downstream tasks. By performing occupancy prediction of the 3D space in the environment, the ability and robustness of scene understanding can be effectively improved. However, existing occupancy prediction tasks are primarily modeled using voxel or point cloud-based approaches: voxel-based network structures often suffer from the loss of spatial information due to the voxelization process, while point cloud-based methods, although better at retaining spatial location information, face limitations in representing volumetric structural details. To address this issue, we propose a dual-modal prediction method based on 3D Gaussian sets and sparse points, which balances both spatial location and volumetric structural information, achieving higher accuracy in semantic occupancy prediction. Specifically, our method adopts a Transformer-based architecture, taking 3D Gaussian sets, sparse points, and queries as inputs. Through the multi-layer structure of the Transformer, the enhanced queries and 3D Gaussian sets jointly contribute to the semantic occupancy prediction, and an adaptive fusion mechanism integrates the semantic outputs of both modalities to generate the final prediction results. Additionally, to further improve accuracy, we dynamically refine the point cloud at each layer, allowing for more precise location information during occupancy prediction. We conducted experiments on the Occ3DnuScenes dataset, and the experimental results demonstrate superior performance of the proposed method on IoU based metrics.
2503.09950
Yuxiang Fu
Yuxiang Fu, Qi Yan, Lele Wang, Ke Li, Renjie Liao
MoFlow: One-Step Flow Matching for Human Trajectory Forecasting via Implicit Maximum Likelihood Estimation based Distillation
Accepted to CVPR 2025
null
null
null
cs.CV cs.AI cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper, we address the problem of human trajectory forecasting, which aims to predict the inherently multi-modal future movements of humans based on their past trajectories and other contextual cues. We propose a novel motion prediction conditional flow matching model, termed MoFlow, to predict K-shot future trajectories for all agents in a given scene. We design a novel flow matching loss function that not only ensures at least one of the $K$ sets of future trajectories is accurate but also encourages all $K$ sets of future trajectories to be diverse and plausible. Furthermore, by leveraging the implicit maximum likelihood estimation (IMLE), we propose a novel distillation method for flow models that only requires samples from the teacher model. Extensive experiments on the real-world datasets, including SportVU NBA games, ETH-UCY, and SDD, demonstrate that both our teacher flow model and the IMLE-distilled student model achieve state-of-the-art performance. These models can generate diverse trajectories that are physically and socially plausible. Moreover, our one-step student model is $\textbf{100}$ times faster than the teacher flow model during sampling. The code, model, and data are available at our project page: https://moflow-imle.github.io
[ { "version": "v1", "created": "Thu, 13 Mar 2025 01:53:05 GMT" } ]
2025-03-14T00:00:00
[ [ "Fu", "Yuxiang", "" ], [ "Yan", "Qi", "" ], [ "Wang", "Lele", "" ], [ "Li", "Ke", "" ], [ "Liao", "Renjie", "" ] ]
TITLE: MoFlow: One-Step Flow Matching for Human Trajectory Forecasting via Implicit Maximum Likelihood Estimation based Distillation ABSTRACT: In this paper, we address the problem of human trajectory forecasting, which aims to predict the inherently multi-modal future movements of humans based on their past trajectories and other contextual cues. We propose a novel motion prediction conditional flow matching model, termed MoFlow, to predict K-shot future trajectories for all agents in a given scene. We design a novel flow matching loss function that not only ensures at least one of the $K$ sets of future trajectories is accurate but also encourages all $K$ sets of future trajectories to be diverse and plausible. Furthermore, by leveraging the implicit maximum likelihood estimation (IMLE), we propose a novel distillation method for flow models that only requires samples from the teacher model. Extensive experiments on the real-world datasets, including SportVU NBA games, ETH-UCY, and SDD, demonstrate that both our teacher flow model and the IMLE-distilled student model achieve state-of-the-art performance. These models can generate diverse trajectories that are physically and socially plausible. Moreover, our one-step student model is $\textbf{100}$ times faster than the teacher flow model during sampling. The code, model, and data are available at our project page: https://moflow-imle.github.io
2503.09959
Jiansheng Li
Jiansheng Li, Haotian Song, Jinni Zhou, Qiang Nie and Yi Cai
RMG: Real-Time Expressive Motion Generation with Self-collision Avoidance for 6-DOF Companion Robotic Arms
null
null
null
null
cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The six-degree-of-freedom (6-DOF) robotic arm has gained widespread application in human-coexisting environments. While previous research has predominantly focused on functional motion generation, the critical aspect of expressive motion in human-robot interaction remains largely unexplored. This paper presents a novel real-time motion generation planner that enhances interactivity by creating expressive robotic motions between arbitrary start and end states within predefined time constraints. Our approach involves three key contributions: first, we develop a mapping algorithm to construct an expressive motion dataset derived from human dance movements; second, we train motion generation models in both Cartesian and joint spaces using this dataset; third, we introduce an optimization algorithm that guarantees smooth, collision-free motion while maintaining the intended expressive style. Experimental results demonstrate the effectiveness of our method, which can generate expressive and generalized motions in under 0.5 seconds while satisfying all specified constraints.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 02:02:01 GMT" } ]
2025-03-14T00:00:00
[ [ "Li", "Jiansheng", "" ], [ "Song", "Haotian", "" ], [ "Zhou", "Jinni", "" ], [ "Nie", "Qiang", "" ], [ "Cai", "Yi", "" ] ]
TITLE: RMG: Real-Time Expressive Motion Generation with Self-collision Avoidance for 6-DOF Companion Robotic Arms ABSTRACT: The six-degree-of-freedom (6-DOF) robotic arm has gained widespread application in human-coexisting environments. While previous research has predominantly focused on functional motion generation, the critical aspect of expressive motion in human-robot interaction remains largely unexplored. This paper presents a novel real-time motion generation planner that enhances interactivity by creating expressive robotic motions between arbitrary start and end states within predefined time constraints. Our approach involves three key contributions: first, we develop a mapping algorithm to construct an expressive motion dataset derived from human dance movements; second, we train motion generation models in both Cartesian and joint spaces using this dataset; third, we introduce an optimization algorithm that guarantees smooth, collision-free motion while maintaining the intended expressive style. Experimental results demonstrate the effectiveness of our method, which can generate expressive and generalized motions in under 0.5 seconds while satisfying all specified constraints.
2503.09960
Muhammad Shahbaz Khan
Muhammad Hassan Jamal, Abdulwahab Alazeb, Shahid Allah Bakhsh, Wadii Boulila, Syed Aziz Shah, Aizaz Ahmad Khattak and Muhammad Shahbaz Khan
Optimizing Fire Safety: Reducing False Alarms Using Advanced Machine Learning Techniques
null
null
null
null
cs.LG cs.AI
http://creativecommons.org/licenses/by-nc-nd/4.0/
Fire safety practices are important to reduce the extent of destruction caused by fire. While smoke alarms help save lives, firefighters struggle with the increasing number of false alarms. This paper presents a precise and efficient Weighted ensemble model for decreasing false alarms. It estimates the density, computes weights according to the high and low-density regions, forwards the high region weights to KNN and low region weights to XGBoost and combines the predictions. The proposed model is effective at reducing response time, increasing fire safety, and minimizing the damage that fires cause. A specifically designed dataset for smoke detection is utilized to test the proposed model. In addition, a variety of ML models, such as Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Nai:ve Bayes (NB), K-Nearest Neighbour (KNN), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Adaptive Boosting (ADAB), have also been utilized. To maximize the use of the smoke detection dataset, all the algorithms utilize the SMOTE re-sampling technique. After evaluating the assessment criteria, this paper presents a concise summary of the comprehensive findings obtained by comparing the outcomes of all models.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 02:07:14 GMT" } ]
2025-03-14T00:00:00
[ [ "Jamal", "Muhammad Hassan", "" ], [ "Alazeb", "Abdulwahab", "" ], [ "Bakhsh", "Shahid Allah", "" ], [ "Boulila", "Wadii", "" ], [ "Shah", "Syed Aziz", "" ], [ "Khattak", "Aizaz Ahmad", "" ], [ "Khan", "Muhammad...
TITLE: Optimizing Fire Safety: Reducing False Alarms Using Advanced Machine Learning Techniques ABSTRACT: Fire safety practices are important to reduce the extent of destruction caused by fire. While smoke alarms help save lives, firefighters struggle with the increasing number of false alarms. This paper presents a precise and efficient Weighted ensemble model for decreasing false alarms. It estimates the density, computes weights according to the high and low-density regions, forwards the high region weights to KNN and low region weights to XGBoost and combines the predictions. The proposed model is effective at reducing response time, increasing fire safety, and minimizing the damage that fires cause. A specifically designed dataset for smoke detection is utilized to test the proposed model. In addition, a variety of ML models, such as Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Nai:ve Bayes (NB), K-Nearest Neighbour (KNN), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Adaptive Boosting (ADAB), have also been utilized. To maximize the use of the smoke detection dataset, all the algorithms utilize the SMOTE re-sampling technique. After evaluating the assessment criteria, this paper presents a concise summary of the comprehensive findings obtained by comparing the outcomes of all models.
2503.09964
Usman Naseem
Bhavik Chandna, Mariam Aboujenane, Usman Naseem
ExtremeAIGC: Benchmarking LMM Vulnerability to AI-Generated Extremist Content
Preprint
null
null
null
cs.CR cs.CL
http://creativecommons.org/licenses/by/4.0/
Large Multimodal Models (LMMs) are increasingly vulnerable to AI-generated extremist content, including photorealistic images and text, which can be used to bypass safety mechanisms and generate harmful outputs. However, existing datasets for evaluating LMM robustness offer limited exploration of extremist content, often lacking AI-generated images, diverse image generation models, and comprehensive coverage of historical events, which hinders a complete assessment of model vulnerabilities. To fill this gap, we introduce ExtremeAIGC, a benchmark dataset and evaluation framework designed to assess LMM vulnerabilities against such content. ExtremeAIGC simulates real-world events and malicious use cases by curating diverse text- and image-based examples crafted using state-of-the-art image generation techniques. Our study reveals alarming weaknesses in LMMs, demonstrating that even cutting-edge safety measures fail to prevent the generation of extremist material. We systematically quantify the success rates of various attack strategies, exposing critical gaps in current defenses and emphasizing the need for more robust mitigation strategies.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 02:10:29 GMT" } ]
2025-03-14T00:00:00
[ [ "Chandna", "Bhavik", "" ], [ "Aboujenane", "Mariam", "" ], [ "Naseem", "Usman", "" ] ]
TITLE: ExtremeAIGC: Benchmarking LMM Vulnerability to AI-Generated Extremist Content ABSTRACT: Large Multimodal Models (LMMs) are increasingly vulnerable to AI-generated extremist content, including photorealistic images and text, which can be used to bypass safety mechanisms and generate harmful outputs. However, existing datasets for evaluating LMM robustness offer limited exploration of extremist content, often lacking AI-generated images, diverse image generation models, and comprehensive coverage of historical events, which hinders a complete assessment of model vulnerabilities. To fill this gap, we introduce ExtremeAIGC, a benchmark dataset and evaluation framework designed to assess LMM vulnerabilities against such content. ExtremeAIGC simulates real-world events and malicious use cases by curating diverse text- and image-based examples crafted using state-of-the-art image generation techniques. Our study reveals alarming weaknesses in LMMs, demonstrating that even cutting-edge safety measures fail to prevent the generation of extremist material. We systematically quantify the success rates of various attack strategies, exposing critical gaps in current defenses and emphasizing the need for more robust mitigation strategies.
2503.09969
Nathan Drenkow
Nathan Drenkow and Mitchell Pavlak and Keith Harrigian and Ayah Zirikly and Adarsh Subbaswamy and Mathias Unberath
Detecting Dataset Bias in Medical AI: A Generalized and Modality-Agnostic Auditing Framework
null
null
null
null
cs.LG cs.AI cs.CV
http://creativecommons.org/licenses/by/4.0/
Data-driven AI is establishing itself at the center of evidence-based medicine. However, reports of shortcomings and unexpected behavior are growing due to AI's reliance on association-based learning. A major reason for this behavior: latent bias in machine learning datasets can be amplified during training and/or hidden during testing. We present a data modality-agnostic auditing framework for generating targeted hypotheses about sources of bias which we refer to as Generalized Attribute Utility and Detectability-Induced bias Testing (G-AUDIT) for datasets. Our method examines the relationship between task-level annotations and data properties including protected attributes (e.g., race, age, sex) and environment and acquisition characteristics (e.g., clinical site, imaging protocols). G-AUDIT automatically quantifies the extent to which the observed data attributes may enable shortcut learning, or in the case of testing data, hide predictions made based on spurious associations. We demonstrate the broad applicability and value of our method by analyzing large-scale medical datasets for three distinct modalities and learning tasks: skin lesion classification in images, stigmatizing language classification in Electronic Health Records (EHR), and mortality prediction for ICU tabular data. In each setting, G-AUDIT successfully identifies subtle biases commonly overlooked by traditional qualitative methods that focus primarily on social and ethical objectives, underscoring its practical value in exposing dataset-level risks and supporting the downstream development of reliable AI systems. Our method paves the way for achieving deeper understanding of machine learning datasets throughout the AI development life-cycle from initial prototyping all the way to regulation, and creates opportunities to reduce model bias, enabling safer and more trustworthy AI systems.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 02:16:48 GMT" } ]
2025-03-14T00:00:00
[ [ "Drenkow", "Nathan", "" ], [ "Pavlak", "Mitchell", "" ], [ "Harrigian", "Keith", "" ], [ "Zirikly", "Ayah", "" ], [ "Subbaswamy", "Adarsh", "" ], [ "Unberath", "Mathias", "" ] ]
TITLE: Detecting Dataset Bias in Medical AI: A Generalized and Modality-Agnostic Auditing Framework ABSTRACT: Data-driven AI is establishing itself at the center of evidence-based medicine. However, reports of shortcomings and unexpected behavior are growing due to AI's reliance on association-based learning. A major reason for this behavior: latent bias in machine learning datasets can be amplified during training and/or hidden during testing. We present a data modality-agnostic auditing framework for generating targeted hypotheses about sources of bias which we refer to as Generalized Attribute Utility and Detectability-Induced bias Testing (G-AUDIT) for datasets. Our method examines the relationship between task-level annotations and data properties including protected attributes (e.g., race, age, sex) and environment and acquisition characteristics (e.g., clinical site, imaging protocols). G-AUDIT automatically quantifies the extent to which the observed data attributes may enable shortcut learning, or in the case of testing data, hide predictions made based on spurious associations. We demonstrate the broad applicability and value of our method by analyzing large-scale medical datasets for three distinct modalities and learning tasks: skin lesion classification in images, stigmatizing language classification in Electronic Health Records (EHR), and mortality prediction for ICU tabular data. In each setting, G-AUDIT successfully identifies subtle biases commonly overlooked by traditional qualitative methods that focus primarily on social and ethical objectives, underscoring its practical value in exposing dataset-level risks and supporting the downstream development of reliable AI systems. Our method paves the way for achieving deeper understanding of machine learning datasets throughout the AI development life-cycle from initial prototyping all the way to regulation, and creates opportunities to reduce model bias, enabling safer and more trustworthy AI systems.
2503.09974
Jiaqi Wu
Jiaqi Wu, Junbiao Pang, Qingming Huang
Uncertainty-aware Long-tailed Weights Model the Utility of Pseudo-labels for Semi-supervised Learning
arXiv admin note: text overlap with arXiv:2408.04150
null
null
null
cs.CV cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Current Semi-supervised Learning (SSL) adopts the pseudo-labeling strategy and further filters pseudo-labels based on confidence thresholds. However, this mechanism has notable drawbacks: 1) setting the reasonable threshold is an open problem which significantly influences the selection of the high-quality pseudo-labels; and 2) deep models often exhibit the over-confidence phenomenon which makes the confidence value an unreliable indicator for assessing the quality of pseudo-labels due to the scarcity of labeled data. In this paper, we propose an Uncertainty-aware Ensemble Structure (UES) to assess the utility of pseudo-labels for unlabeled samples. We further model the utility of pseudo-labels as long-tailed weights to avoid the open problem of setting the threshold. Concretely, the advantage of the long-tailed weights ensures that even unreliable pseudo-labels still contribute to enhancing the model's robustness. Besides, UES is lightweight and architecture-agnostic, easily extending to various computer vision tasks, including classification and regression. Experimental results demonstrate that combining the proposed method with DualPose leads to a 3.47% improvement in Percentage of Correct Keypoints (PCK) on the Sniffing dataset with 100 data points (30 labeled), a 7.29\% improvement in PCK on the FLIC dataset with 100 data points (50 labeled), and a 3.91% improvement in PCK on the LSP dataset with 200 data points (100 labeled). Furthermore, when combined with FixMatch, the proposed method achieves a 0.2% accuracy improvement on the CIFAR-10 dataset with 40 labeled data points and a 0.26% accuracy improvement on the CIFAR-100 dataset with 400 labeled data points.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 02:21:04 GMT" } ]
2025-03-14T00:00:00
[ [ "Wu", "Jiaqi", "" ], [ "Pang", "Junbiao", "" ], [ "Huang", "Qingming", "" ] ]
TITLE: Uncertainty-aware Long-tailed Weights Model the Utility of Pseudo-labels for Semi-supervised Learning ABSTRACT: Current Semi-supervised Learning (SSL) adopts the pseudo-labeling strategy and further filters pseudo-labels based on confidence thresholds. However, this mechanism has notable drawbacks: 1) setting the reasonable threshold is an open problem which significantly influences the selection of the high-quality pseudo-labels; and 2) deep models often exhibit the over-confidence phenomenon which makes the confidence value an unreliable indicator for assessing the quality of pseudo-labels due to the scarcity of labeled data. In this paper, we propose an Uncertainty-aware Ensemble Structure (UES) to assess the utility of pseudo-labels for unlabeled samples. We further model the utility of pseudo-labels as long-tailed weights to avoid the open problem of setting the threshold. Concretely, the advantage of the long-tailed weights ensures that even unreliable pseudo-labels still contribute to enhancing the model's robustness. Besides, UES is lightweight and architecture-agnostic, easily extending to various computer vision tasks, including classification and regression. Experimental results demonstrate that combining the proposed method with DualPose leads to a 3.47% improvement in Percentage of Correct Keypoints (PCK) on the Sniffing dataset with 100 data points (30 labeled), a 7.29\% improvement in PCK on the FLIC dataset with 100 data points (50 labeled), and a 3.91% improvement in PCK on the LSP dataset with 200 data points (100 labeled). Furthermore, when combined with FixMatch, the proposed method achieves a 0.2% accuracy improvement on the CIFAR-10 dataset with 40 labeled data points and a 0.26% accuracy improvement on the CIFAR-100 dataset with 400 labeled data points.
2503.09978
Jiacheng Xie
Jiacheng Xie, Hua-Chieh Shao, Yunxiang Li, Shunyu Yan, Chenyang Shen, Jing Wang, You Zhang
A Conditional Point Cloud Diffusion Model for Deformable Liver Motion Tracking Via a Single Arbitrarily-Angled X-ray Projection
25 pages, 7 figures
null
null
null
physics.med-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deformable liver motion tracking using a single X-ray projection enables real-time motion monitoring and treatment intervention. We introduce a conditional point cloud diffusion model-based framework for accurate and robust liver motion tracking from arbitrarily angled single X-ray projections (PCD-Liver), which estimates volumetric liver motion by solving deformable vector fields (DVFs) of a prior liver surface point cloud based on a single X-ray image. The model is patient-specific and consists of two main components: a rigid alignment model to estimate the liver's overall shifts and a conditional point cloud diffusion model that further corrects for liver surface deformations. Conditioned on motion-encoded features extracted from a single X-ray projection via a geometry-informed feature pooling layer, the diffusion model iteratively solves detailed liver surface DVFs in a projection angle-agnostic manner. The liver surface motion estimated by PCD-Liver serves as a boundary condition for a U-Net-based biomechanical model to infer internal liver motion and localize liver tumors. A dataset of ten liver cancer patients was used for evaluation. The accuracy of liver point cloud motion estimation was assessed using root mean square error (RMSE) and 95th-percentile Hausdorff distance (HD95), while liver tumor localization error was quantified using center-of-mass error (COME). The mean (standard deviation) RMSE, HD95, and COME of the prior liver or tumor before motion estimation were 8.86(1.51) mm, 10.88(2.56) mm, and 9.41(3.08) mm, respectively. After PCD-Liver motion estimation, the corresponding values improved to 3.59(0.28) mm, 4.29(0.62) mm, and 3.45(0.96) mm. Under highly noisy conditions, PCD-Liver maintained stable performance. This study presents an accurate and robust framework for deformable liver motion estimation and tumor localization in image-guided radiotherapy.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 02:27:26 GMT" } ]
2025-03-14T00:00:00
[ [ "Xie", "Jiacheng", "" ], [ "Shao", "Hua-Chieh", "" ], [ "Li", "Yunxiang", "" ], [ "Yan", "Shunyu", "" ], [ "Shen", "Chenyang", "" ], [ "Wang", "Jing", "" ], [ "Zhang", "You", "" ] ]
TITLE: A Conditional Point Cloud Diffusion Model for Deformable Liver Motion Tracking Via a Single Arbitrarily-Angled X-ray Projection ABSTRACT: Deformable liver motion tracking using a single X-ray projection enables real-time motion monitoring and treatment intervention. We introduce a conditional point cloud diffusion model-based framework for accurate and robust liver motion tracking from arbitrarily angled single X-ray projections (PCD-Liver), which estimates volumetric liver motion by solving deformable vector fields (DVFs) of a prior liver surface point cloud based on a single X-ray image. The model is patient-specific and consists of two main components: a rigid alignment model to estimate the liver's overall shifts and a conditional point cloud diffusion model that further corrects for liver surface deformations. Conditioned on motion-encoded features extracted from a single X-ray projection via a geometry-informed feature pooling layer, the diffusion model iteratively solves detailed liver surface DVFs in a projection angle-agnostic manner. The liver surface motion estimated by PCD-Liver serves as a boundary condition for a U-Net-based biomechanical model to infer internal liver motion and localize liver tumors. A dataset of ten liver cancer patients was used for evaluation. The accuracy of liver point cloud motion estimation was assessed using root mean square error (RMSE) and 95th-percentile Hausdorff distance (HD95), while liver tumor localization error was quantified using center-of-mass error (COME). The mean (standard deviation) RMSE, HD95, and COME of the prior liver or tumor before motion estimation were 8.86(1.51) mm, 10.88(2.56) mm, and 9.41(3.08) mm, respectively. After PCD-Liver motion estimation, the corresponding values improved to 3.59(0.28) mm, 4.29(0.62) mm, and 3.45(0.96) mm. Under highly noisy conditions, PCD-Liver maintained stable performance. This study presents an accurate and robust framework for deformable liver motion estimation and tumor localization in image-guided radiotherapy.
2503.09994
Yunxiao Wang
Yunxiao Wang, Meng Liu, Rui Shao, Haoyu Zhang, Bin Wen, Fan Yang, Tingting Gao, Di Zhang, Liqiang Nie
TIME: Temporal-sensitive Multi-dimensional Instruction Tuning and Benchmarking for Video-LLMs
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Video large language models have achieved remarkable performance in tasks such as video question answering, however, their temporal understanding remains suboptimal. To address this limitation, we curate a dedicated instruction fine-tuning dataset that focuses on enhancing temporal comprehension across five key dimensions. In order to reduce reliance on costly temporal annotations, we introduce a multi-task prompt fine-tuning approach that seamlessly integrates temporal-sensitive tasks into existing instruction datasets without requiring additional annotations. Furthermore, we develop a novel benchmark for temporal-sensitive video understanding that not only fills the gaps in dimension coverage left by existing benchmarks but also rigorously filters out potential shortcuts, ensuring a more accurate evaluation. Extensive experimental results demonstrate that our approach significantly enhances the temporal understanding of video-LLMs while avoiding reliance on shortcuts.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 03:05:11 GMT" } ]
2025-03-14T00:00:00
[ [ "Wang", "Yunxiao", "" ], [ "Liu", "Meng", "" ], [ "Shao", "Rui", "" ], [ "Zhang", "Haoyu", "" ], [ "Wen", "Bin", "" ], [ "Yang", "Fan", "" ], [ "Gao", "Tingting", "" ], [ "Zhang", "Di", "" ...
TITLE: TIME: Temporal-sensitive Multi-dimensional Instruction Tuning and Benchmarking for Video-LLMs ABSTRACT: Video large language models have achieved remarkable performance in tasks such as video question answering, however, their temporal understanding remains suboptimal. To address this limitation, we curate a dedicated instruction fine-tuning dataset that focuses on enhancing temporal comprehension across five key dimensions. In order to reduce reliance on costly temporal annotations, we introduce a multi-task prompt fine-tuning approach that seamlessly integrates temporal-sensitive tasks into existing instruction datasets without requiring additional annotations. Furthermore, we develop a novel benchmark for temporal-sensitive video understanding that not only fills the gaps in dimension coverage left by existing benchmarks but also rigorously filters out potential shortcuts, ensuring a more accurate evaluation. Extensive experimental results demonstrate that our approach significantly enhances the temporal understanding of video-LLMs while avoiding reliance on shortcuts.
2503.10009
Bowen Zhang
Bowen Zhang, Pengcheng Luo
OR-LLM-Agent: Automating Modeling and Solving of Operations Research Optimization Problem with Reasoning Large Language Model
11 pages, 6 figures
null
null
null
cs.AI math.OC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Operations Research (OR) has been widely applied in various fields such as resource allocation, production planning, and supply chain management. However, addressing real-world OR problems requires OR experts to perform mathematical modeling and programmers to develop solution algorithms. This traditional method, heavily reliant on experts, is costly and has long development cycles, severely limiting the widespread adoption of OR techniques. Few have considered using Artificial Intelligence (AI) to replace professionals to achieve fully automated solutions for OR problems. We propose OR-LLM-Agent, the first AI agent that enables end-to-end automation for solving real-world OR problems. OR-LLM-Agent leverages the Chain-of-Thought (CoT) reasoning capabilities of Large Language Models (LLMs) to translate natural language problem descriptions into formal mathematical models and automatically generate Gurobi solver code. In OR-LLM-Agent, OR-CodeAgent is designed to automate code execution and repair within a sandbox environment, facilitating the derivation of the final solution. Due to the lack of dedicated benchmark datasets for evaluating the automated solving of OR problems, we construct a benchmark dataset comprising 83 real-world OR problems described in natural language. We conduct comparative experiments with state-of-the-art (SOTA) reasoning LLMs, including GPT-o3-mini, DeepSeek-R1, and Gemini 2.0 Flash Thinking. The OR-LLM-Agent achieved the highest pass rate of 100% and the highest solution accuracy of 85%, demonstrating the feasibility of automated OR problem-solving. Data and code have been publicly available at https://github.com/bwz96sco/or_llm_agent.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 03:40:50 GMT" } ]
2025-03-14T00:00:00
[ [ "Zhang", "Bowen", "" ], [ "Luo", "Pengcheng", "" ] ]
TITLE: OR-LLM-Agent: Automating Modeling and Solving of Operations Research Optimization Problem with Reasoning Large Language Model ABSTRACT: Operations Research (OR) has been widely applied in various fields such as resource allocation, production planning, and supply chain management. However, addressing real-world OR problems requires OR experts to perform mathematical modeling and programmers to develop solution algorithms. This traditional method, heavily reliant on experts, is costly and has long development cycles, severely limiting the widespread adoption of OR techniques. Few have considered using Artificial Intelligence (AI) to replace professionals to achieve fully automated solutions for OR problems. We propose OR-LLM-Agent, the first AI agent that enables end-to-end automation for solving real-world OR problems. OR-LLM-Agent leverages the Chain-of-Thought (CoT) reasoning capabilities of Large Language Models (LLMs) to translate natural language problem descriptions into formal mathematical models and automatically generate Gurobi solver code. In OR-LLM-Agent, OR-CodeAgent is designed to automate code execution and repair within a sandbox environment, facilitating the derivation of the final solution. Due to the lack of dedicated benchmark datasets for evaluating the automated solving of OR problems, we construct a benchmark dataset comprising 83 real-world OR problems described in natural language. We conduct comparative experiments with state-of-the-art (SOTA) reasoning LLMs, including GPT-o3-mini, DeepSeek-R1, and Gemini 2.0 Flash Thinking. The OR-LLM-Agent achieved the highest pass rate of 100% and the highest solution accuracy of 85%, demonstrating the feasibility of automated OR problem-solving. Data and code have been publicly available at https://github.com/bwz96sco/or_llm_agent.
2503.10034
Hao Xiang
Hao Xiang, Zhaoliang Zheng, Xin Xia, Seth Z. Zhao, Letian Gao, Zewei Zhou, Tianhui Cai, Yun Zhang, Jiaqi Ma
V2X-ReaLO: An Open Online Framework and Dataset for Cooperative Perception in Reality
null
null
null
null
cs.CV cs.RO
http://creativecommons.org/licenses/by-nc-nd/4.0/
Cooperative perception enabled by Vehicle-to-Everything (V2X) communication holds significant promise for enhancing the perception capabilities of autonomous vehicles, allowing them to overcome occlusions and extend their field of view. However, existing research predominantly relies on simulated environments or static datasets, leaving the feasibility and effectiveness of V2X cooperative perception especially for intermediate fusion in real-world scenarios largely unexplored. In this work, we introduce V2X-ReaLO, an open online cooperative perception framework deployed on real vehicles and smart infrastructure that integrates early, late, and intermediate fusion methods within a unified pipeline and provides the first practical demonstration of online intermediate fusion's feasibility and performance under genuine real-world conditions. Additionally, we present an open benchmark dataset specifically designed to assess the performance of online cooperative perception systems. This new dataset extends V2X-Real dataset to dynamic, synchronized ROS bags and provides 25,028 test frames with 6,850 annotated key frames in challenging urban scenarios. By enabling real-time assessments of perception accuracy and communication lantency under dynamic conditions, V2X-ReaLO sets a new benchmark for advancing and optimizing cooperative perception systems in real-world applications. The codes and datasets will be released to further advance the field.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 04:31:20 GMT" } ]
2025-03-14T00:00:00
[ [ "Xiang", "Hao", "" ], [ "Zheng", "Zhaoliang", "" ], [ "Xia", "Xin", "" ], [ "Zhao", "Seth Z.", "" ], [ "Gao", "Letian", "" ], [ "Zhou", "Zewei", "" ], [ "Cai", "Tianhui", "" ], [ "Zhang", "Yun",...
TITLE: V2X-ReaLO: An Open Online Framework and Dataset for Cooperative Perception in Reality ABSTRACT: Cooperative perception enabled by Vehicle-to-Everything (V2X) communication holds significant promise for enhancing the perception capabilities of autonomous vehicles, allowing them to overcome occlusions and extend their field of view. However, existing research predominantly relies on simulated environments or static datasets, leaving the feasibility and effectiveness of V2X cooperative perception especially for intermediate fusion in real-world scenarios largely unexplored. In this work, we introduce V2X-ReaLO, an open online cooperative perception framework deployed on real vehicles and smart infrastructure that integrates early, late, and intermediate fusion methods within a unified pipeline and provides the first practical demonstration of online intermediate fusion's feasibility and performance under genuine real-world conditions. Additionally, we present an open benchmark dataset specifically designed to assess the performance of online cooperative perception systems. This new dataset extends V2X-Real dataset to dynamic, synchronized ROS bags and provides 25,028 test frames with 6,850 annotated key frames in challenging urban scenarios. By enabling real-time assessments of perception accuracy and communication lantency under dynamic conditions, V2X-ReaLO sets a new benchmark for advancing and optimizing cooperative perception systems in real-world applications. The codes and datasets will be released to further advance the field.
2503.10040
Seunghun Lee
Dongik Lee, Valentin Stanev, Xiaohang Zhang, Mijeong Kang, Ichiro Takeuchi, and Seunghun Lee
Rapid analysis of point-contact Andreev reflection spectra via machine learning with adaptive data augmentation
18 pages, 3 figures
null
null
null
cond-mat.supr-con cs.AI cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Delineating the superconducting order parameters is a pivotal task in investigating superconductivity for probing pairing mechanisms, as well as their symmetry and topology. Point-contact Andreev reflection (PCAR) measurement is a simple yet powerful tool for identifying the order parameters. The PCAR spectra exhibit significant variations depending on the type of the order parameter in a superconductor, including its magnitude ($\mathit{\Delta}$), as well as temperature, interfacial quality, Fermi velocity mismatch, and other factors. The information on the order parameter can be obtained by finding the combination of these parameters, generating a theoretical spectrum that fits a measured experimental spectrum. However, due to the complexity of the spectra and the high dimensionality of parameters, extracting the fitting parameters is often time-consuming and labor-intensive. In this study, we employ a convolutional neural network (CNN) algorithm to create models for rapid and automated analysis of PCAR spectra of various superconductors with different pairing symmetries (conventional $s$-wave, chiral $p_x+ip_y$-wave, and $d_{x^2-y^2}$-wave). The training datasets are generated based on the Blonder-Tinkham-Klapwijk (BTK) theory and further modified and augmented by selectively incorporating noise and peaks according to the bias voltages. This approach not only replicates the experimental spectra but also brings the model's attention to important features within the spectra. The optimized models provide fitting parameters for experimentally measured spectra in less than 100 ms per spectrum. Our approaches and findings pave the way for rapid and automated spectral analysis which will help accelerate research on superconductors with complex order parameters.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 04:45:38 GMT" } ]
2025-03-14T00:00:00
[ [ "Lee", "Dongik", "" ], [ "Stanev", "Valentin", "" ], [ "Zhang", "Xiaohang", "" ], [ "Kang", "Mijeong", "" ], [ "Takeuchi", "Ichiro", "" ], [ "Lee", "Seunghun", "" ] ]
TITLE: Rapid analysis of point-contact Andreev reflection spectra via machine learning with adaptive data augmentation ABSTRACT: Delineating the superconducting order parameters is a pivotal task in investigating superconductivity for probing pairing mechanisms, as well as their symmetry and topology. Point-contact Andreev reflection (PCAR) measurement is a simple yet powerful tool for identifying the order parameters. The PCAR spectra exhibit significant variations depending on the type of the order parameter in a superconductor, including its magnitude ($\mathit{\Delta}$), as well as temperature, interfacial quality, Fermi velocity mismatch, and other factors. The information on the order parameter can be obtained by finding the combination of these parameters, generating a theoretical spectrum that fits a measured experimental spectrum. However, due to the complexity of the spectra and the high dimensionality of parameters, extracting the fitting parameters is often time-consuming and labor-intensive. In this study, we employ a convolutional neural network (CNN) algorithm to create models for rapid and automated analysis of PCAR spectra of various superconductors with different pairing symmetries (conventional $s$-wave, chiral $p_x+ip_y$-wave, and $d_{x^2-y^2}$-wave). The training datasets are generated based on the Blonder-Tinkham-Klapwijk (BTK) theory and further modified and augmented by selectively incorporating noise and peaks according to the bias voltages. This approach not only replicates the experimental spectra but also brings the model's attention to important features within the spectra. The optimized models provide fitting parameters for experimentally measured spectra in less than 100 ms per spectrum. Our approaches and findings pave the way for rapid and automated spectral analysis which will help accelerate research on superconductors with complex order parameters.
2503.10045
Yaoting Jiang
Meng Wang, Zi Yang, Ruifeng Zhao, Yaoting Jiang
CPLOYO: A Pulmonary Nodule Detection Model with Multi-Scale Feature Fusion and Nonlinear Feature Learning
null
null
null
null
eess.IV cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The integration of Internet of Things (IoT) technology in pulmonary nodule detection significantly enhances the intelligence and real-time capabilities of the detection system. Currently, lung nodule detection primarily focuses on the identification of solid nodules, but different types of lung nodules correspond to various forms of lung cancer. Multi-type detection contributes to improving the overall lung cancer detection rate and enhancing the cure rate. To achieve high sensitivity in nodule detection, targeted improvements were made to the YOLOv8 model. Firstly, the C2f\_RepViTCAMF module was introduced to augment the C2f module in the backbone, thereby enhancing detection accuracy for small lung nodules and achieving a lightweight model design. Secondly, the MSCAF module was incorporated to reconstruct the feature fusion section of the model, improving detection accuracy for lung nodules of varying scales. Furthermore, the KAN network was integrated into the model. By leveraging the KAN network's powerful nonlinear feature learning capability, detection accuracy for small lung nodules was further improved, and the model's generalization ability was enhanced. Tests conducted on the LUNA16 dataset demonstrate that the improved model outperforms the original model as well as other mainstream models such as YOLOv9 and RT-DETR across various evaluation metrics.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 04:51:57 GMT" } ]
2025-03-14T00:00:00
[ [ "Wang", "Meng", "" ], [ "Yang", "Zi", "" ], [ "Zhao", "Ruifeng", "" ], [ "Jiang", "Yaoting", "" ] ]
TITLE: CPLOYO: A Pulmonary Nodule Detection Model with Multi-Scale Feature Fusion and Nonlinear Feature Learning ABSTRACT: The integration of Internet of Things (IoT) technology in pulmonary nodule detection significantly enhances the intelligence and real-time capabilities of the detection system. Currently, lung nodule detection primarily focuses on the identification of solid nodules, but different types of lung nodules correspond to various forms of lung cancer. Multi-type detection contributes to improving the overall lung cancer detection rate and enhancing the cure rate. To achieve high sensitivity in nodule detection, targeted improvements were made to the YOLOv8 model. Firstly, the C2f\_RepViTCAMF module was introduced to augment the C2f module in the backbone, thereby enhancing detection accuracy for small lung nodules and achieving a lightweight model design. Secondly, the MSCAF module was incorporated to reconstruct the feature fusion section of the model, improving detection accuracy for lung nodules of varying scales. Furthermore, the KAN network was integrated into the model. By leveraging the KAN network's powerful nonlinear feature learning capability, detection accuracy for small lung nodules was further improved, and the model's generalization ability was enhanced. Tests conducted on the LUNA16 dataset demonstrate that the improved model outperforms the original model as well as other mainstream models such as YOLOv9 and RT-DETR across various evaluation metrics.
2503.10052
Minjun Kim
Minje Kim, Minjun Kim, Xu Yang
DTA: Dual Temporal-channel-wise Attention for Spiking Neural Networks
Accepted by IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2025
null
null
null
cs.CV cs.AI cs.NE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Spiking Neural Networks (SNNs) present a more energy-efficient alternative to Artificial Neural Networks (ANNs) by harnessing spatio-temporal dynamics and event-driven spikes. Effective utilization of temporal information is crucial for SNNs, leading to the exploration of attention mechanisms to enhance this capability. Conventional attention operations either apply identical operation or employ non-identical operations across target dimensions. We identify that these approaches provide distinct perspectives on temporal information. To leverage the strengths of both operations, we propose a novel Dual Temporal-channel-wise Attention (DTA) mechanism that integrates both identical/non-identical attention strategies. To the best of our knowledge, this is the first attempt to concentrate on both the correlation and dependency of temporal-channel using both identical and non-identical attention operations. Experimental results demonstrate that the DTA mechanism achieves state-of-the-art performance on both static datasets (CIFAR10, CIFAR100, ImageNet-1k) and dynamic dataset (CIFAR10-DVS), elevating spike representation and capturing complex temporal-channel relationship. We open-source our code: https://github.com/MnJnKIM/DTA-SNN.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 05:09:48 GMT" } ]
2025-03-14T00:00:00
[ [ "Kim", "Minje", "" ], [ "Kim", "Minjun", "" ], [ "Yang", "Xu", "" ] ]
TITLE: DTA: Dual Temporal-channel-wise Attention for Spiking Neural Networks ABSTRACT: Spiking Neural Networks (SNNs) present a more energy-efficient alternative to Artificial Neural Networks (ANNs) by harnessing spatio-temporal dynamics and event-driven spikes. Effective utilization of temporal information is crucial for SNNs, leading to the exploration of attention mechanisms to enhance this capability. Conventional attention operations either apply identical operation or employ non-identical operations across target dimensions. We identify that these approaches provide distinct perspectives on temporal information. To leverage the strengths of both operations, we propose a novel Dual Temporal-channel-wise Attention (DTA) mechanism that integrates both identical/non-identical attention strategies. To the best of our knowledge, this is the first attempt to concentrate on both the correlation and dependency of temporal-channel using both identical and non-identical attention operations. Experimental results demonstrate that the DTA mechanism achieves state-of-the-art performance on both static datasets (CIFAR10, CIFAR100, ImageNet-1k) and dynamic dataset (CIFAR10-DVS), elevating spike representation and capturing complex temporal-channel relationship. We open-source our code: https://github.com/MnJnKIM/DTA-SNN.
2503.10055
Donghyun Kim
Donghyun Kim, Hyunah Ko, Chanyoung Kim, Seong Jae Hwang
Fourier Decomposition for Explicit Representation of 3D Point Cloud Attributes
null
null
null
null
cs.CV eess.IV
http://creativecommons.org/licenses/by-nc-nd/4.0/
While 3D point clouds are widely utilized across various vision applications, their irregular and sparse nature make them challenging to handle. In response, numerous encoding approaches have been proposed to capture the rich semantic information of point clouds. Yet, a critical limitation persists: a lack of consideration for colored point clouds which are more capable 3D representations as they contain diverse attributes: color and geometry. While existing methods handle these attributes separately on a per-point basis, this leads to a limited receptive field and restricted ability to capture relationships across multiple points. To address this, we pioneer a point cloud encoding methodology that leverages 3D Fourier decomposition to disentangle color and geometric features while extending the receptive field through spectral-domain operations. Our analysis confirms that this encoding approach effectively separates feature components, where the amplitude uniquely captures color attributes and the phase encodes geometric structure, thereby enabling independent learning and utilization of both attributes. Furthermore, the spectral-domain properties of these components naturally aggregate local features while considering multiple points' information. We validate our point cloud encoding approach on point cloud classification and style transfer tasks, achieving state-of-the-art results on the DensePoint dataset with improvements via a proposed amplitude-based data augmentation strategy.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 05:13:40 GMT" } ]
2025-03-14T00:00:00
[ [ "Kim", "Donghyun", "" ], [ "Ko", "Hyunah", "" ], [ "Kim", "Chanyoung", "" ], [ "Hwang", "Seong Jae", "" ] ]
TITLE: Fourier Decomposition for Explicit Representation of 3D Point Cloud Attributes ABSTRACT: While 3D point clouds are widely utilized across various vision applications, their irregular and sparse nature make them challenging to handle. In response, numerous encoding approaches have been proposed to capture the rich semantic information of point clouds. Yet, a critical limitation persists: a lack of consideration for colored point clouds which are more capable 3D representations as they contain diverse attributes: color and geometry. While existing methods handle these attributes separately on a per-point basis, this leads to a limited receptive field and restricted ability to capture relationships across multiple points. To address this, we pioneer a point cloud encoding methodology that leverages 3D Fourier decomposition to disentangle color and geometric features while extending the receptive field through spectral-domain operations. Our analysis confirms that this encoding approach effectively separates feature components, where the amplitude uniquely captures color attributes and the phase encodes geometric structure, thereby enabling independent learning and utilization of both attributes. Furthermore, the spectral-domain properties of these components naturally aggregate local features while considering multiple points' information. We validate our point cloud encoding approach on point cloud classification and style transfer tasks, achieving state-of-the-art results on the DensePoint dataset with improvements via a proposed amplitude-based data augmentation strategy.
2503.10057
Ho Hin Lee
Ho Hin Lee, Alberto Santamaria-Pang, Jameson Merkov, Matthew Lungren, Ivan Tarapov
Multi-Modal Mamba Modeling for Survival Prediction (M4Survive): Adapting Joint Foundation Model Representations
10 pages
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
Accurate survival prediction in oncology requires integrating diverse imaging modalities to capture the complex interplay of tumor biology. Traditional single-modality approaches often fail to leverage the complementary insights provided by radiological and pathological assessments. In this work, we introduce M4Survive (Multi-Modal Mamba Modeling for Survival Prediction), a novel framework that learns joint foundation model representations using efficient adapter networks. Our approach dynamically fuses heterogeneous embeddings from a foundation model repository (e.g., MedImageInsight, BiomedCLIP, Prov-GigaPath, UNI2-h), creating a correlated latent space optimized for survival risk estimation. By leveraging Mamba-based adapters, M4Survive enables efficient multi-modal learning while preserving computational efficiency. Experimental evaluations on benchmark datasets demonstrate that our approach outperforms both unimodal and traditional static multi-modal baselines in survival prediction accuracy. This work underscores the potential of foundation model-driven multi-modal fusion in advancing precision oncology and predictive analytics.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 05:18:32 GMT" } ]
2025-03-14T00:00:00
[ [ "Lee", "Ho Hin", "" ], [ "Santamaria-Pang", "Alberto", "" ], [ "Merkov", "Jameson", "" ], [ "Lungren", "Matthew", "" ], [ "Tarapov", "Ivan", "" ] ]
TITLE: Multi-Modal Mamba Modeling for Survival Prediction (M4Survive): Adapting Joint Foundation Model Representations ABSTRACT: Accurate survival prediction in oncology requires integrating diverse imaging modalities to capture the complex interplay of tumor biology. Traditional single-modality approaches often fail to leverage the complementary insights provided by radiological and pathological assessments. In this work, we introduce M4Survive (Multi-Modal Mamba Modeling for Survival Prediction), a novel framework that learns joint foundation model representations using efficient adapter networks. Our approach dynamically fuses heterogeneous embeddings from a foundation model repository (e.g., MedImageInsight, BiomedCLIP, Prov-GigaPath, UNI2-h), creating a correlated latent space optimized for survival risk estimation. By leveraging Mamba-based adapters, M4Survive enables efficient multi-modal learning while preserving computational efficiency. Experimental evaluations on benchmark datasets demonstrate that our approach outperforms both unimodal and traditional static multi-modal baselines in survival prediction accuracy. This work underscores the potential of foundation model-driven multi-modal fusion in advancing precision oncology and predictive analytics.
2503.10071
Sunzida Siddique
Mohd Ariful Haque, Justin Williams, Sunzida Siddique, Md. Hujaifa Islam, Hasmot Ali, Kishor Datta Gupta, and Roy George
Advanced Tool Learning and Selection System (ATLASS): A Closed-Loop Framework Using LLM
null
null
null
null
cs.AI
http://creativecommons.org/licenses/by-sa/4.0/
The combination of LLM agents with external tools enables models to solve complex tasks beyond their knowledge base. Human-designed tools are inflexible and restricted to solutions within the scope of pre-existing tools created by experts. To address this problem, we propose ATLASS, an advanced tool learning and selection system designed as a closed-loop framework. It enables the LLM to solve problems by dynamically generating external tools on demand. In this framework, agents play a crucial role in orchestrating tool selection, execution, and refinement, ensuring adaptive problem-solving capabilities. The operation of ATLASS follows three phases: The first phase, Understanding Tool Requirements, involves the Agents determining whether tools are required and specifying their functionality; the second phase, Tool Retrieval/Generation, involves the Agents retrieving or generating tools based on their availability; and the third phase, Task Solving, involves combining all the component tools necessary to complete the initial task. The Tool Dataset stores the generated tools, ensuring reusability and minimizing inference cost. Current LLM-based tool generation systems have difficulty creating complex tools that need APIs or external packages. In ATLASS, we solve the problem by automatically setting up the environment, fetching relevant API documentation online, and using a Python interpreter to create a reliable, versatile tool that works in a wider range of situations. OpenAI GPT-4.0 is used as the LLM agent, and safety and ethical concerns are handled through human feedback before executing generated code. By addressing the limitations of predefined toolsets and enhancing adaptability, ATLASS serves as a real-world solution that empowers users with dynamically generated tools for complex problem-solving.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 05:39:00 GMT" } ]
2025-03-14T00:00:00
[ [ "Haque", "Mohd Ariful", "" ], [ "Williams", "Justin", "" ], [ "Siddique", "Sunzida", "" ], [ "Islam", "Md. Hujaifa", "" ], [ "Ali", "Hasmot", "" ], [ "Gupta", "Kishor Datta", "" ], [ "George", "Roy", "" ]...
TITLE: Advanced Tool Learning and Selection System (ATLASS): A Closed-Loop Framework Using LLM ABSTRACT: The combination of LLM agents with external tools enables models to solve complex tasks beyond their knowledge base. Human-designed tools are inflexible and restricted to solutions within the scope of pre-existing tools created by experts. To address this problem, we propose ATLASS, an advanced tool learning and selection system designed as a closed-loop framework. It enables the LLM to solve problems by dynamically generating external tools on demand. In this framework, agents play a crucial role in orchestrating tool selection, execution, and refinement, ensuring adaptive problem-solving capabilities. The operation of ATLASS follows three phases: The first phase, Understanding Tool Requirements, involves the Agents determining whether tools are required and specifying their functionality; the second phase, Tool Retrieval/Generation, involves the Agents retrieving or generating tools based on their availability; and the third phase, Task Solving, involves combining all the component tools necessary to complete the initial task. The Tool Dataset stores the generated tools, ensuring reusability and minimizing inference cost. Current LLM-based tool generation systems have difficulty creating complex tools that need APIs or external packages. In ATLASS, we solve the problem by automatically setting up the environment, fetching relevant API documentation online, and using a Python interpreter to create a reliable, versatile tool that works in a wider range of situations. OpenAI GPT-4.0 is used as the LLM agent, and safety and ethical concerns are handled through human feedback before executing generated code. By addressing the limitations of predefined toolsets and enhancing adaptability, ATLASS serves as a real-world solution that empowers users with dynamically generated tools for complex problem-solving.
2503.10092
Xintong Dong
Xintong Dong, Wenshuo Yu, Jun Lin, Zhenbo Guo, Hongzhou Wang, Jianhao Yang
Light-weighted foundation model for seismic data processing based on representative and non-redundant pre-training dataset
null
null
null
null
physics.geo-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In the fields of computer vision (CV) and remote sensing (RS), foundational models typically follow the "big data + large model parameters" paradigm. However, the application of this strategy in seismic data processing faces several challenges: seismic data is difficult to obtain and the scarcity of publicly available datasets make it difficult to construct large-scale datasets. Additionally, the high computational cost associated with a large number of model parameters restricts widespread research in this domain. Therefore, we propose a lightweight seismic processing foundational model paradigm (SPFM), which aims to overcome the limitations of traditional methods by data engineering and network architecture innovation. Specifically, we propose an innovative dataset construction strategy that generates more seismic data by data augmentation techniques, including collecting publicly available field data and using generative diffusion models (GDM) for data enhancement. Furthermore, we optimize the data distribution by employing dimensionality reduction, cluster analysis, and stratified sampling methods, reducing redundant information while preserving important seismic features, thus constructing a comprehensive dataset. In terms of network architecture design, we introduce the selective structured state-space model (Mamba) structure, which effectively captures global features of seismic data and alleviates the quadratic growth of computational complexity inherent in Transformer-based models, thereby improving computational efficiency. This model, pre-trained with only four A800 GPUs, outperforms traditional methods across multiple tasks, including denoising, interpolation, frequency-band extrapolation, and resolution enhancement. The lightweight paradigm provides an solution for seismic data processing, advancing the generalization and accessibility of seismic data processing.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 06:40:33 GMT" } ]
2025-03-14T00:00:00
[ [ "Dong", "Xintong", "" ], [ "Yu", "Wenshuo", "" ], [ "Lin", "Jun", "" ], [ "Guo", "Zhenbo", "" ], [ "Wang", "Hongzhou", "" ], [ "Yang", "Jianhao", "" ] ]
TITLE: Light-weighted foundation model for seismic data processing based on representative and non-redundant pre-training dataset ABSTRACT: In the fields of computer vision (CV) and remote sensing (RS), foundational models typically follow the "big data + large model parameters" paradigm. However, the application of this strategy in seismic data processing faces several challenges: seismic data is difficult to obtain and the scarcity of publicly available datasets make it difficult to construct large-scale datasets. Additionally, the high computational cost associated with a large number of model parameters restricts widespread research in this domain. Therefore, we propose a lightweight seismic processing foundational model paradigm (SPFM), which aims to overcome the limitations of traditional methods by data engineering and network architecture innovation. Specifically, we propose an innovative dataset construction strategy that generates more seismic data by data augmentation techniques, including collecting publicly available field data and using generative diffusion models (GDM) for data enhancement. Furthermore, we optimize the data distribution by employing dimensionality reduction, cluster analysis, and stratified sampling methods, reducing redundant information while preserving important seismic features, thus constructing a comprehensive dataset. In terms of network architecture design, we introduce the selective structured state-space model (Mamba) structure, which effectively captures global features of seismic data and alleviates the quadratic growth of computational complexity inherent in Transformer-based models, thereby improving computational efficiency. This model, pre-trained with only four A800 GPUs, outperforms traditional methods across multiple tasks, including denoising, interpolation, frequency-band extrapolation, and resolution enhancement. The lightweight paradigm provides an solution for seismic data processing, advancing the generalization and accessibility of seismic data processing.
2503.10115
Hanlin Pan
Hanlin Pan, Kunpeng Liu, Wanfu Gao
Reconsidering Feature Structure Information and Latent Space Alignment in Partial Multi-label Feature Selection
9pages,6 figures,accept at AAAI 25
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The purpose of partial multi-label feature selection is to select the most representative feature subset, where the data comes from partial multi-label datasets that have label ambiguity issues. For label disambiguation, previous methods mainly focus on utilizing the information inside the labels and the relationship between the labels and features. However, the information existing in the feature space is rarely considered, especially in partial multi-label scenarios where the noises is considered to be concentrated in the label space while the feature information is correct. This paper proposes a method based on latent space alignment, which uses the information mined in feature space to disambiguate in latent space through the structural consistency between labels and features. In addition, previous methods overestimate the consistency of features and labels in the latent space after convergence. We comprehensively consider the similarity of latent space projections to feature space and label space, and propose new feature selection term. This method also significantly improves the positive label identification ability of the selected features. Comprehensive experiments demonstrate the superiority of the proposed method.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 07:21:29 GMT" } ]
2025-03-14T00:00:00
[ [ "Pan", "Hanlin", "" ], [ "Liu", "Kunpeng", "" ], [ "Gao", "Wanfu", "" ] ]
TITLE: Reconsidering Feature Structure Information and Latent Space Alignment in Partial Multi-label Feature Selection ABSTRACT: The purpose of partial multi-label feature selection is to select the most representative feature subset, where the data comes from partial multi-label datasets that have label ambiguity issues. For label disambiguation, previous methods mainly focus on utilizing the information inside the labels and the relationship between the labels and features. However, the information existing in the feature space is rarely considered, especially in partial multi-label scenarios where the noises is considered to be concentrated in the label space while the feature information is correct. This paper proposes a method based on latent space alignment, which uses the information mined in feature space to disambiguate in latent space through the structural consistency between labels and features. In addition, previous methods overestimate the consistency of features and labels in the latent space after convergence. We comprehensively consider the similarity of latent space projections to feature space and label space, and propose new feature selection term. This method also significantly improves the positive label identification ability of the selected features. Comprehensive experiments demonstrate the superiority of the proposed method.
2503.10120
Bingchen Li
Bingchen Li, Xin Li, Yiting Lu, Zhibo Chen
Hybrid Agents for Image Restoration
null
null
null
null
cs.CV eess.IV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Existing Image Restoration (IR) studies typically focus on task-specific or universal modes individually, relying on the mode selection of users and lacking the cooperation between multiple task-specific/universal restoration modes. This leads to insufficient interaction for unprofessional users and limits their restoration capability for complicated real-world applications. In this work, we present HybridAgent, intending to incorporate multiple restoration modes into a unified image restoration model and achieve intelligent and efficient user interaction through our proposed hybrid agents. Concretely, we propose the hybrid rule of fast, slow, and feedback restoration agents. Here, the slow restoration agent optimizes the powerful multimodal large language model (MLLM) with our proposed instruction-tuning dataset to identify degradations within images with ambiguous user prompts and invokes proper restoration tools accordingly. The fast restoration agent is designed based on a lightweight large language model (LLM) via in-context learning to understand the user prompts with simple and clear requirements, which can obviate the unnecessary time/resource costs of MLLM. Moreover, we introduce the mixed distortion removal mode for our HybridAgents, which is crucial but not concerned in previous agent-based works. It can effectively prevent the error propagation of step-by-step image restoration and largely improve the efficiency of the agent system. We validate the effectiveness of HybridAgent with both synthetic and real-world IR tasks.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 07:28:33 GMT" } ]
2025-03-14T00:00:00
[ [ "Li", "Bingchen", "" ], [ "Li", "Xin", "" ], [ "Lu", "Yiting", "" ], [ "Chen", "Zhibo", "" ] ]
TITLE: Hybrid Agents for Image Restoration ABSTRACT: Existing Image Restoration (IR) studies typically focus on task-specific or universal modes individually, relying on the mode selection of users and lacking the cooperation between multiple task-specific/universal restoration modes. This leads to insufficient interaction for unprofessional users and limits their restoration capability for complicated real-world applications. In this work, we present HybridAgent, intending to incorporate multiple restoration modes into a unified image restoration model and achieve intelligent and efficient user interaction through our proposed hybrid agents. Concretely, we propose the hybrid rule of fast, slow, and feedback restoration agents. Here, the slow restoration agent optimizes the powerful multimodal large language model (MLLM) with our proposed instruction-tuning dataset to identify degradations within images with ambiguous user prompts and invokes proper restoration tools accordingly. The fast restoration agent is designed based on a lightweight large language model (LLM) via in-context learning to understand the user prompts with simple and clear requirements, which can obviate the unnecessary time/resource costs of MLLM. Moreover, we introduce the mixed distortion removal mode for our HybridAgents, which is crucial but not concerned in previous agent-based works. It can effectively prevent the error propagation of step-by-step image restoration and largely improve the efficiency of the agent system. We validate the effectiveness of HybridAgent with both synthetic and real-world IR tasks.
2503.10129
Namal Jayasuriya
Namal Jayasuriya, Yi Guo, Wen Hu, Oula Ghannoum
Deep Learning-Based Direct Leaf Area Estimation using Two RGBD Datasets for Model Development
null
null
null
null
cs.CV cs.AI
http://creativecommons.org/licenses/by-nc-nd/4.0/
Estimation of a single leaf area can be a measure of crop growth and a phenotypic trait to breed new varieties. It has also been used to measure leaf area index and total leaf area. Some studies have used hand-held cameras, image processing 3D reconstruction and unsupervised learning-based methods to estimate the leaf area in plant images. Deep learning works well for object detection and segmentation tasks; however, direct area estimation of objects has not been explored. This work investigates deep learning-based leaf area estimation, for RGBD images taken using a mobile camera setup in real-world scenarios. A dataset for attached leaves captured with a top angle view and a dataset for detached single leaves were collected for model development and testing. First, image processing-based area estimation was tested on manually segmented leaves. Then a Mask R-CNN-based model was investigated, and modified to accept RGBD images and to estimate the leaf area. The detached-leaf data set was then mixed with the attached-leaf plant data set to estimate the single leaf area for plant images, and another network design with two backbones was proposed: one for segmentation and the other for area estimation. Instead of trying all possibilities or random values, an agile approach was used in hyperparameter tuning. The final model was cross-validated with 5-folds and tested with two unseen datasets: detached and attached leaves. The F1 score with 90% IoA for segmentation result on unseen detached-leaf data was 1.0, while R-squared of area estimation was 0.81. For unseen plant data segmentation, the F1 score with 90% IoA was 0.59, while the R-squared score was 0.57. The research suggests using attached leaves with ground truth area to improve the results.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 07:39:09 GMT" } ]
2025-03-14T00:00:00
[ [ "Jayasuriya", "Namal", "" ], [ "Guo", "Yi", "" ], [ "Hu", "Wen", "" ], [ "Ghannoum", "Oula", "" ] ]
TITLE: Deep Learning-Based Direct Leaf Area Estimation using Two RGBD Datasets for Model Development ABSTRACT: Estimation of a single leaf area can be a measure of crop growth and a phenotypic trait to breed new varieties. It has also been used to measure leaf area index and total leaf area. Some studies have used hand-held cameras, image processing 3D reconstruction and unsupervised learning-based methods to estimate the leaf area in plant images. Deep learning works well for object detection and segmentation tasks; however, direct area estimation of objects has not been explored. This work investigates deep learning-based leaf area estimation, for RGBD images taken using a mobile camera setup in real-world scenarios. A dataset for attached leaves captured with a top angle view and a dataset for detached single leaves were collected for model development and testing. First, image processing-based area estimation was tested on manually segmented leaves. Then a Mask R-CNN-based model was investigated, and modified to accept RGBD images and to estimate the leaf area. The detached-leaf data set was then mixed with the attached-leaf plant data set to estimate the single leaf area for plant images, and another network design with two backbones was proposed: one for segmentation and the other for area estimation. Instead of trying all possibilities or random values, an agile approach was used in hyperparameter tuning. The final model was cross-validated with 5-folds and tested with two unseen datasets: detached and attached leaves. The F1 score with 90% IoA for segmentation result on unseen detached-leaf data was 1.0, while R-squared of area estimation was 0.81. For unseen plant data segmentation, the F1 score with 90% IoA was 0.59, while the R-squared score was 0.57. The research suggests using attached leaves with ground truth area to improve the results.
2503.10149
Zhenxuan Zeng
Zhenxuan Zeng, Qiao Wu, Xiyu Zhang, Lin Yuanbo Wu, Pei An, Jiaqi Yang, Ji Wang, Peng Wang
Unlocking Generalization Power in LiDAR Point Cloud Registration
Accepted by CVPR 2025
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In real-world environments, a LiDAR point cloud registration method with robust generalization capabilities (across varying distances and datasets) is crucial for ensuring safety in autonomous driving and other LiDAR-based applications. However, current methods fall short in achieving this level of generalization. To address these limitations, we propose UGP, a pruned framework designed to enhance generalization power for LiDAR point cloud registration. The core insight in UGP is the elimination of cross-attention mechanisms to improve generalization, allowing the network to concentrate on intra-frame feature extraction. Additionally, we introduce a progressive self-attention module to reduce ambiguity in large-scale scenes and integrate Bird's Eye View (BEV) features to incorporate semantic information about scene elements. Together, these enhancements significantly boost the network's generalization performance. We validated our approach through various generalization experiments in multiple outdoor scenes. In cross-distance generalization experiments on KITTI and nuScenes, UGP achieved state-of-the-art mean Registration Recall rates of 94.5% and 91.4%, respectively. In cross-dataset generalization from nuScenes to KITTI, UGP achieved a state-of-the-art mean Registration Recall of 90.9%. Code will be available at https://github.com/peakpang/UGP.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 08:20:59 GMT" } ]
2025-03-14T00:00:00
[ [ "Zeng", "Zhenxuan", "" ], [ "Wu", "Qiao", "" ], [ "Zhang", "Xiyu", "" ], [ "Wu", "Lin Yuanbo", "" ], [ "An", "Pei", "" ], [ "Yang", "Jiaqi", "" ], [ "Wang", "Ji", "" ], [ "Wang", "Peng", "" ...
TITLE: Unlocking Generalization Power in LiDAR Point Cloud Registration ABSTRACT: In real-world environments, a LiDAR point cloud registration method with robust generalization capabilities (across varying distances and datasets) is crucial for ensuring safety in autonomous driving and other LiDAR-based applications. However, current methods fall short in achieving this level of generalization. To address these limitations, we propose UGP, a pruned framework designed to enhance generalization power for LiDAR point cloud registration. The core insight in UGP is the elimination of cross-attention mechanisms to improve generalization, allowing the network to concentrate on intra-frame feature extraction. Additionally, we introduce a progressive self-attention module to reduce ambiguity in large-scale scenes and integrate Bird's Eye View (BEV) features to incorporate semantic information about scene elements. Together, these enhancements significantly boost the network's generalization performance. We validated our approach through various generalization experiments in multiple outdoor scenes. In cross-distance generalization experiments on KITTI and nuScenes, UGP achieved state-of-the-art mean Registration Recall rates of 94.5% and 91.4%, respectively. In cross-dataset generalization from nuScenes to KITTI, UGP achieved a state-of-the-art mean Registration Recall of 90.9%. Code will be available at https://github.com/peakpang/UGP.
2503.10152
Shenghao Fu
Shenghao Fu, Junkai Yan, Qize Yang, Xihan Wei, Xiaohua Xie, Wei-Shi Zheng
A Hierarchical Semantic Distillation Framework for Open-Vocabulary Object Detection
Accepted to TMM 2025
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Open-vocabulary object detection (OVD) aims to detect objects beyond the training annotations, where detectors are usually aligned to a pre-trained vision-language model, eg, CLIP, to inherit its generalizable recognition ability so that detectors can recognize new or novel objects. However, previous works directly align the feature space with CLIP and fail to learn the semantic knowledge effectively. In this work, we propose a hierarchical semantic distillation framework named HD-OVD to construct a comprehensive distillation process, which exploits generalizable knowledge from the CLIP model in three aspects. In the first hierarchy of HD-OVD, the detector learns fine-grained instance-wise semantics from the CLIP image encoder by modeling relations among single objects in the visual space. Besides, we introduce text space novel-class-aware classification to help the detector assimilate the highly generalizable class-wise semantics from the CLIP text encoder, representing the second hierarchy. Lastly, abundant image-wise semantics containing multi-object and their contexts are also distilled by an image-wise contrastive distillation. Benefiting from the elaborated semantic distillation in triple hierarchies, our HD-OVD inherits generalizable recognition ability from CLIP in instance, class, and image levels. Thus, we boost the novel AP on the OV-COCO dataset to 46.4% with a ResNet50 backbone, which outperforms others by a clear margin. We also conduct extensive ablation studies to analyze how each component works.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 08:27:18 GMT" } ]
2025-03-14T00:00:00
[ [ "Fu", "Shenghao", "" ], [ "Yan", "Junkai", "" ], [ "Yang", "Qize", "" ], [ "Wei", "Xihan", "" ], [ "Xie", "Xiaohua", "" ], [ "Zheng", "Wei-Shi", "" ] ]
TITLE: A Hierarchical Semantic Distillation Framework for Open-Vocabulary Object Detection ABSTRACT: Open-vocabulary object detection (OVD) aims to detect objects beyond the training annotations, where detectors are usually aligned to a pre-trained vision-language model, eg, CLIP, to inherit its generalizable recognition ability so that detectors can recognize new or novel objects. However, previous works directly align the feature space with CLIP and fail to learn the semantic knowledge effectively. In this work, we propose a hierarchical semantic distillation framework named HD-OVD to construct a comprehensive distillation process, which exploits generalizable knowledge from the CLIP model in three aspects. In the first hierarchy of HD-OVD, the detector learns fine-grained instance-wise semantics from the CLIP image encoder by modeling relations among single objects in the visual space. Besides, we introduce text space novel-class-aware classification to help the detector assimilate the highly generalizable class-wise semantics from the CLIP text encoder, representing the second hierarchy. Lastly, abundant image-wise semantics containing multi-object and their contexts are also distilled by an image-wise contrastive distillation. Benefiting from the elaborated semantic distillation in triple hierarchies, our HD-OVD inherits generalizable recognition ability from CLIP in instance, class, and image levels. Thus, we boost the novel AP on the OV-COCO dataset to 46.4% with a ResNet50 backbone, which outperforms others by a clear margin. We also conduct extensive ablation studies to analyze how each component works.
2503.10154
Junghyo Jo
Yechan Lim, Sangwon Lee, Junghyo Jo
Data augmentation using diffusion models to enhance inverse Ising inference
null
null
null
null
physics.data-an cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Identifying model parameters from observed configurations poses a fundamental challenge in data science, especially with limited data. Recently, diffusion models have emerged as a novel paradigm in generative machine learning, capable of producing new samples that closely mimic observed data. These models learn the gradient of model probabilities, bypassing the need for cumbersome calculations of partition functions across all possible configurations. We explore whether diffusion models can enhance parameter inference by augmenting small datasets. Our findings demonstrate this potential through a synthetic task involving inverse Ising inference and a real-world application of reconstructing missing values in neural activity data. This study serves as a proof-of-concept for using diffusion models for data augmentation in physics-related problems, thereby opening new avenues in data science.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 08:29:17 GMT" } ]
2025-03-14T00:00:00
[ [ "Lim", "Yechan", "" ], [ "Lee", "Sangwon", "" ], [ "Jo", "Junghyo", "" ] ]
TITLE: Data augmentation using diffusion models to enhance inverse Ising inference ABSTRACT: Identifying model parameters from observed configurations poses a fundamental challenge in data science, especially with limited data. Recently, diffusion models have emerged as a novel paradigm in generative machine learning, capable of producing new samples that closely mimic observed data. These models learn the gradient of model probabilities, bypassing the need for cumbersome calculations of partition functions across all possible configurations. We explore whether diffusion models can enhance parameter inference by augmenting small datasets. Our findings demonstrate this potential through a synthetic task involving inverse Ising inference and a real-world application of reconstructing missing values in neural activity data. This study serves as a proof-of-concept for using diffusion models for data augmentation in physics-related problems, thereby opening new avenues in data science.
2503.10166
Pengfei Luo
Pengfei Luo, Jingbo Zhou, Tong Xu, Yuan Xia, Linli Xu, Enhong Chen
ImageScope: Unifying Language-Guided Image Retrieval via Large Multimodal Model Collective Reasoning
WWW 2025
null
null
null
cs.IR cs.AI cs.MM
http://creativecommons.org/licenses/by-nc-nd/4.0/
With the proliferation of images in online content, language-guided image retrieval (LGIR) has emerged as a research hotspot over the past decade, encompassing a variety of subtasks with diverse input forms. While the development of large multimodal models (LMMs) has significantly facilitated these tasks, existing approaches often address them in isolation, requiring the construction of separate systems for each task. This not only increases system complexity and maintenance costs, but also exacerbates challenges stemming from language ambiguity and complex image content, making it difficult for retrieval systems to provide accurate and reliable results. To this end, we propose ImageScope, a training-free, three-stage framework that leverages collective reasoning to unify LGIR tasks. The key insight behind the unification lies in the compositional nature of language, which transforms diverse LGIR tasks into a generalized text-to-image retrieval process, along with the reasoning of LMMs serving as a universal verification to refine the results. To be specific, in the first stage, we improve the robustness of the framework by synthesizing search intents across varying levels of semantic granularity using chain-of-thought (CoT) reasoning. In the second and third stages, we then reflect on retrieval results by verifying predicate propositions locally, and performing pairwise evaluations globally. Experiments conducted on six LGIR datasets demonstrate that ImageScope outperforms competitive baselines. Comprehensive evaluations and ablation studies further confirm the effectiveness of our design.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 08:43:24 GMT" } ]
2025-03-14T00:00:00
[ [ "Luo", "Pengfei", "" ], [ "Zhou", "Jingbo", "" ], [ "Xu", "Tong", "" ], [ "Xia", "Yuan", "" ], [ "Xu", "Linli", "" ], [ "Chen", "Enhong", "" ] ]
TITLE: ImageScope: Unifying Language-Guided Image Retrieval via Large Multimodal Model Collective Reasoning ABSTRACT: With the proliferation of images in online content, language-guided image retrieval (LGIR) has emerged as a research hotspot over the past decade, encompassing a variety of subtasks with diverse input forms. While the development of large multimodal models (LMMs) has significantly facilitated these tasks, existing approaches often address them in isolation, requiring the construction of separate systems for each task. This not only increases system complexity and maintenance costs, but also exacerbates challenges stemming from language ambiguity and complex image content, making it difficult for retrieval systems to provide accurate and reliable results. To this end, we propose ImageScope, a training-free, three-stage framework that leverages collective reasoning to unify LGIR tasks. The key insight behind the unification lies in the compositional nature of language, which transforms diverse LGIR tasks into a generalized text-to-image retrieval process, along with the reasoning of LMMs serving as a universal verification to refine the results. To be specific, in the first stage, we improve the robustness of the framework by synthesizing search intents across varying levels of semantic granularity using chain-of-thought (CoT) reasoning. In the second and third stages, we then reflect on retrieval results by verifying predicate propositions locally, and performing pairwise evaluations globally. Experiments conducted on six LGIR datasets demonstrate that ImageScope outperforms competitive baselines. Comprehensive evaluations and ablation studies further confirm the effectiveness of our design.
2503.10195
Daqing Guo
Hongze Sun, Jun Wang, Wuque Cai, Duo Chen, Qianqian Liao, Jiayi He, Yan Cui, Dezhong Yao, Daqing Guo
ST-FlowNet: An Efficient Spiking Neural Network for Event-Based Optical Flow Estimation
12 pages, 5 figures, 5 tables; This work has been submitted for possible publication
null
null
null
cs.CV cs.NE q-bio.NC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Spiking Neural Networks (SNNs) have emerged as a promising tool for event-based optical flow estimation tasks due to their ability to leverage spatio-temporal information and low-power capabilities. However, the performance of SNN models is often constrained, limiting their application in real-world scenarios. In this work, we address this gap by proposing a novel neural network architecture, ST-FlowNet, specifically tailored for optical flow estimation from event-based data. The ST-FlowNet architecture integrates ConvGRU modules to facilitate cross-modal feature augmentation and temporal alignment of the predicted optical flow, improving the network's ability to capture complex motion dynamics. Additionally, to overcome the challenges associated with training SNNs, we introduce a novel approach to derive SNN models from pre-trained artificial neural networks (ANNs) through ANN-to-SNN conversion or our proposed BISNN method. Notably, the BISNN method alleviates the complexities involved in biological parameter selection, further enhancing the robustness of SNNs in optical flow estimation tasks. Extensive evaluations on three benchmark event-based datasets demonstrate that the SNN-based ST-FlowNet model outperforms state-of-the-art methods, delivering superior performance in accurate optical flow estimation across a diverse range of dynamic visual scenes. Furthermore, the inherent energy efficiency of SNN models is highlighted, establishing a compelling advantage for their practical deployment. Overall, our work presents a novel framework for optical flow estimation using SNNs and event-based data, contributing to the advancement of neuromorphic vision applications.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 09:28:42 GMT" } ]
2025-03-14T00:00:00
[ [ "Sun", "Hongze", "" ], [ "Wang", "Jun", "" ], [ "Cai", "Wuque", "" ], [ "Chen", "Duo", "" ], [ "Liao", "Qianqian", "" ], [ "He", "Jiayi", "" ], [ "Cui", "Yan", "" ], [ "Yao", "Dezhong", "" ...
TITLE: ST-FlowNet: An Efficient Spiking Neural Network for Event-Based Optical Flow Estimation ABSTRACT: Spiking Neural Networks (SNNs) have emerged as a promising tool for event-based optical flow estimation tasks due to their ability to leverage spatio-temporal information and low-power capabilities. However, the performance of SNN models is often constrained, limiting their application in real-world scenarios. In this work, we address this gap by proposing a novel neural network architecture, ST-FlowNet, specifically tailored for optical flow estimation from event-based data. The ST-FlowNet architecture integrates ConvGRU modules to facilitate cross-modal feature augmentation and temporal alignment of the predicted optical flow, improving the network's ability to capture complex motion dynamics. Additionally, to overcome the challenges associated with training SNNs, we introduce a novel approach to derive SNN models from pre-trained artificial neural networks (ANNs) through ANN-to-SNN conversion or our proposed BISNN method. Notably, the BISNN method alleviates the complexities involved in biological parameter selection, further enhancing the robustness of SNNs in optical flow estimation tasks. Extensive evaluations on three benchmark event-based datasets demonstrate that the SNN-based ST-FlowNet model outperforms state-of-the-art methods, delivering superior performance in accurate optical flow estimation across a diverse range of dynamic visual scenes. Furthermore, the inherent energy efficiency of SNN models is highlighted, establishing a compelling advantage for their practical deployment. Overall, our work presents a novel framework for optical flow estimation using SNNs and event-based data, contributing to the advancement of neuromorphic vision applications.
2503.10198
Xiangjie Kong
Xiangjie Kong, Zhenghao Chen, Weiyao Liu, Kaili Ning, Lechao Zhang, Syauqie Muhammad Marier, Yichen Liu, Yuhao Chen, Feng Xia
Deep Learning for Time Series Forecasting: A Survey
null
Int. J. Mach. Learn. & Cyber. (2025)
10.1007/s13042-025-02560-w
null
cs.LG cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Time series forecasting (TSF) has long been a crucial task in both industry and daily life. Most classical statistical models may have certain limitations when applied to practical scenarios in fields such as energy, healthcare, traffic, meteorology, and economics, especially when high accuracy is required. With the continuous development of deep learning, numerous new models have emerged in the field of time series forecasting in recent years. However, existing surveys have not provided a unified summary of the wide range of model architectures in this field, nor have they given detailed summaries of works in feature extraction and datasets. To address this gap, in this review, we comprehensively study the previous works and summarize the general paradigms of Deep Time Series Forecasting (DTSF) in terms of model architectures. Besides, we take an innovative approach by focusing on the composition of time series and systematically explain important feature extraction methods. Additionally, we provide an overall compilation of datasets from various domains in existing works. Finally, we systematically emphasize the significant challenges faced and future research directions in this field.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 09:32:01 GMT" } ]
2025-03-14T00:00:00
[ [ "Kong", "Xiangjie", "" ], [ "Chen", "Zhenghao", "" ], [ "Liu", "Weiyao", "" ], [ "Ning", "Kaili", "" ], [ "Zhang", "Lechao", "" ], [ "Marier", "Syauqie Muhammad", "" ], [ "Liu", "Yichen", "" ], [ "C...
TITLE: Deep Learning for Time Series Forecasting: A Survey ABSTRACT: Time series forecasting (TSF) has long been a crucial task in both industry and daily life. Most classical statistical models may have certain limitations when applied to practical scenarios in fields such as energy, healthcare, traffic, meteorology, and economics, especially when high accuracy is required. With the continuous development of deep learning, numerous new models have emerged in the field of time series forecasting in recent years. However, existing surveys have not provided a unified summary of the wide range of model architectures in this field, nor have they given detailed summaries of works in feature extraction and datasets. To address this gap, in this review, we comprehensively study the previous works and summarize the general paradigms of Deep Time Series Forecasting (DTSF) in terms of model architectures. Besides, we take an innovative approach by focusing on the composition of time series and systematically explain important feature extraction methods. Additionally, we provide an overall compilation of datasets from various domains in existing works. Finally, we systematically emphasize the significant challenges faced and future research directions in this field.
2503.10210
Jialong Wu
Jialong Wu, Marco Braun, Dominic Spata, Matthias Rottmann
TARS: Traffic-Aware Radar Scene Flow Estimation
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Scene flow provides crucial motion information for autonomous driving. Recent LiDAR scene flow models utilize the rigid-motion assumption at the instance level, assuming objects are rigid bodies. However, these instance-level methods are not suitable for sparse radar point clouds. In this work, we present a novel $\textbf{T}$raffic-$\textbf{A}$ware $\textbf{R}$adar $\textbf{S}$cene flow estimation method, named $\textbf{TARS}$, which utilizes the motion rigidity at the traffic level. To address the challenges in radar scene flow, we perform object detection and scene flow jointly and boost the latter. We incorporate the feature map from the object detector, trained with detection losses, to make radar scene flow aware of the environment and road users. Therefrom, we construct a Traffic Vector Field (TVF) in the feature space, enabling a holistic traffic-level scene understanding in our scene flow branch. When estimating the scene flow, we consider both point-level motion cues from point neighbors and traffic-level consistency of rigid motion within the space. TARS outperforms the state of the art on a proprietary dataset and the View-of-Delft dataset, improving the benchmarks by 23% and 15%, respectively.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 09:54:08 GMT" } ]
2025-03-14T00:00:00
[ [ "Wu", "Jialong", "" ], [ "Braun", "Marco", "" ], [ "Spata", "Dominic", "" ], [ "Rottmann", "Matthias", "" ] ]
TITLE: TARS: Traffic-Aware Radar Scene Flow Estimation ABSTRACT: Scene flow provides crucial motion information for autonomous driving. Recent LiDAR scene flow models utilize the rigid-motion assumption at the instance level, assuming objects are rigid bodies. However, these instance-level methods are not suitable for sparse radar point clouds. In this work, we present a novel $\textbf{T}$raffic-$\textbf{A}$ware $\textbf{R}$adar $\textbf{S}$cene flow estimation method, named $\textbf{TARS}$, which utilizes the motion rigidity at the traffic level. To address the challenges in radar scene flow, we perform object detection and scene flow jointly and boost the latter. We incorporate the feature map from the object detector, trained with detection losses, to make radar scene flow aware of the environment and road users. Therefrom, we construct a Traffic Vector Field (TVF) in the feature space, enabling a holistic traffic-level scene understanding in our scene flow branch. When estimating the scene flow, we consider both point-level motion cues from point neighbors and traffic-level consistency of rigid motion within the space. TARS outperforms the state of the art on a proprietary dataset and the View-of-Delft dataset, improving the benchmarks by 23% and 15%, respectively.
2503.10214
Zhiwu Wang
Zhiwu Wang, Yichen Wu, Renzhen Wang, Haokun Lin, Quanziang Wang, Qian Zhao, Deyu Meng
Singular Value Fine-tuning for Few-Shot Class-Incremental Learning
12 pages, 8 figures
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Class-Incremental Learning (CIL) aims to prevent catastrophic forgetting of previously learned classes while sequentially incorporating new ones. The more challenging Few-shot CIL (FSCIL) setting further complicates this by providing only a limited number of samples for each new class, increasing the risk of overfitting in addition to standard CIL challenges. While catastrophic forgetting has been extensively studied, overfitting in FSCIL, especially with large foundation models, has received less attention. To fill this gap, we propose the Singular Value Fine-tuning for FSCIL (SVFCL) and compared it with existing approaches for adapting foundation models to FSCIL, which primarily build on Parameter Efficient Fine-Tuning (PEFT) methods like prompt tuning and Low-Rank Adaptation (LoRA). Specifically, SVFCL applies singular value decomposition to the foundation model weights, keeping the singular vectors fixed while fine-tuning the singular values for each task, and then merging them. This simple yet effective approach not only alleviates the forgetting problem but also mitigates overfitting more effectively while significantly reducing trainable parameters. Extensive experiments on four benchmark datasets, along with visualizations and ablation studies, validate the effectiveness of SVFCL. The code will be made available.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 09:57:28 GMT" } ]
2025-03-14T00:00:00
[ [ "Wang", "Zhiwu", "" ], [ "Wu", "Yichen", "" ], [ "Wang", "Renzhen", "" ], [ "Lin", "Haokun", "" ], [ "Wang", "Quanziang", "" ], [ "Zhao", "Qian", "" ], [ "Meng", "Deyu", "" ] ]
TITLE: Singular Value Fine-tuning for Few-Shot Class-Incremental Learning ABSTRACT: Class-Incremental Learning (CIL) aims to prevent catastrophic forgetting of previously learned classes while sequentially incorporating new ones. The more challenging Few-shot CIL (FSCIL) setting further complicates this by providing only a limited number of samples for each new class, increasing the risk of overfitting in addition to standard CIL challenges. While catastrophic forgetting has been extensively studied, overfitting in FSCIL, especially with large foundation models, has received less attention. To fill this gap, we propose the Singular Value Fine-tuning for FSCIL (SVFCL) and compared it with existing approaches for adapting foundation models to FSCIL, which primarily build on Parameter Efficient Fine-Tuning (PEFT) methods like prompt tuning and Low-Rank Adaptation (LoRA). Specifically, SVFCL applies singular value decomposition to the foundation model weights, keeping the singular vectors fixed while fine-tuning the singular values for each task, and then merging them. This simple yet effective approach not only alleviates the forgetting problem but also mitigates overfitting more effectively while significantly reducing trainable parameters. Extensive experiments on four benchmark datasets, along with visualizations and ablation studies, validate the effectiveness of SVFCL. The code will be made available.
2503.10216
Kaixiang Yang
Kaixiang Yang, Xin Li, Qiang Li, Zhiwei Wang
CoStoDet-DDPM: Collaborative Training of Stochastic and Deterministic Models Improves Surgical Workflow Anticipation and Recognition
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Anticipating and recognizing surgical workflows are critical for intelligent surgical assistance systems. However, existing methods rely on deterministic decision-making, struggling to generalize across the large anatomical and procedural variations inherent in real-world surgeries.In this paper, we introduce an innovative framework that incorporates stochastic modeling through a denoising diffusion probabilistic model (DDPM) into conventional deterministic learning for surgical workflow analysis. At the heart of our approach is a collaborative co-training paradigm: the DDPM branch captures procedural uncertainties to enrich feature representations, while the task branch focuses on predicting surgical phases and instrument usage.Theoretically, we demonstrate that this mutual refinement mechanism benefits both branches: the DDPM reduces prediction errors in uncertain scenarios, and the task branch directs the DDPM toward clinically meaningful representations. Notably, the DDPM branch is discarded during inference, enabling real-time predictions without sacrificing accuracy.Experiments on the Cholec80 dataset show that for the anticipation task, our method achieves a 16% reduction in eMAE compared to state-of-the-art approaches, and for phase recognition, it improves the Jaccard score by 1.0%. Additionally, on the AutoLaparo dataset, our method achieves a 1.5% improvement in the Jaccard score for phase recognition, while also exhibiting robust generalization to patient-specific variations. Our code and weight are available at https://github.com/kk42yy/CoStoDet-DDPM.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 09:59:05 GMT" } ]
2025-03-14T00:00:00
[ [ "Yang", "Kaixiang", "" ], [ "Li", "Xin", "" ], [ "Li", "Qiang", "" ], [ "Wang", "Zhiwei", "" ] ]
TITLE: CoStoDet-DDPM: Collaborative Training of Stochastic and Deterministic Models Improves Surgical Workflow Anticipation and Recognition ABSTRACT: Anticipating and recognizing surgical workflows are critical for intelligent surgical assistance systems. However, existing methods rely on deterministic decision-making, struggling to generalize across the large anatomical and procedural variations inherent in real-world surgeries.In this paper, we introduce an innovative framework that incorporates stochastic modeling through a denoising diffusion probabilistic model (DDPM) into conventional deterministic learning for surgical workflow analysis. At the heart of our approach is a collaborative co-training paradigm: the DDPM branch captures procedural uncertainties to enrich feature representations, while the task branch focuses on predicting surgical phases and instrument usage.Theoretically, we demonstrate that this mutual refinement mechanism benefits both branches: the DDPM reduces prediction errors in uncertain scenarios, and the task branch directs the DDPM toward clinically meaningful representations. Notably, the DDPM branch is discarded during inference, enabling real-time predictions without sacrificing accuracy.Experiments on the Cholec80 dataset show that for the anticipation task, our method achieves a 16% reduction in eMAE compared to state-of-the-art approaches, and for phase recognition, it improves the Jaccard score by 1.0%. Additionally, on the AutoLaparo dataset, our method achieves a 1.5% improvement in the Jaccard score for phase recognition, while also exhibiting robust generalization to patient-specific variations. Our code and weight are available at https://github.com/kk42yy/CoStoDet-DDPM.
2503.10225
Zhixuan Li
Zhixuan Li, Hyunse Yoon, Sanghoon Lee, Weisi Lin
Unveiling the Invisible: Reasoning Complex Occlusions Amodally with AURA
11 pages, 5 figures, 5 tables
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Amodal segmentation aims to infer the complete shape of occluded objects, even when the occluded region's appearance is unavailable. However, current amodal segmentation methods lack the capability to interact with users through text input and struggle to understand or reason about implicit and complex purposes. While methods like LISA integrate multi-modal large language models (LLMs) with segmentation for reasoning tasks, they are limited to predicting only visible object regions and face challenges in handling complex occlusion scenarios. To address these limitations, we propose a novel task named amodal reasoning segmentation, aiming to predict the complete amodal shape of occluded objects while providing answers with elaborations based on user text input. We develop a generalizable dataset generation pipeline and introduce a new dataset focusing on daily life scenarios, encompassing diverse real-world occlusions. Furthermore, we present AURA (Amodal Understanding and Reasoning Assistant), a novel model with advanced global and spatial-level designs specifically tailored to handle complex occlusions. Extensive experiments validate AURA's effectiveness on the proposed dataset. The code, model, and dataset will be publicly released.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 10:08:18 GMT" } ]
2025-03-14T00:00:00
[ [ "Li", "Zhixuan", "" ], [ "Yoon", "Hyunse", "" ], [ "Lee", "Sanghoon", "" ], [ "Lin", "Weisi", "" ] ]
TITLE: Unveiling the Invisible: Reasoning Complex Occlusions Amodally with AURA ABSTRACT: Amodal segmentation aims to infer the complete shape of occluded objects, even when the occluded region's appearance is unavailable. However, current amodal segmentation methods lack the capability to interact with users through text input and struggle to understand or reason about implicit and complex purposes. While methods like LISA integrate multi-modal large language models (LLMs) with segmentation for reasoning tasks, they are limited to predicting only visible object regions and face challenges in handling complex occlusion scenarios. To address these limitations, we propose a novel task named amodal reasoning segmentation, aiming to predict the complete amodal shape of occluded objects while providing answers with elaborations based on user text input. We develop a generalizable dataset generation pipeline and introduce a new dataset focusing on daily life scenarios, encompassing diverse real-world occlusions. Furthermore, we present AURA (Amodal Understanding and Reasoning Assistant), a novel model with advanced global and spatial-level designs specifically tailored to handle complex occlusions. Extensive experiments validate AURA's effectiveness on the proposed dataset. The code, model, and dataset will be publicly released.
2503.10228
Andi Nika
Andi Nika, Jonathan N\"other, Debmalya Mandal, Parameswaran Kamalaruban, Adish Singla and Goran Radanovi\'c
Policy Teaching via Data Poisoning in Learning from Human Preferences
In AISTATS 2025
null
null
null
cs.LG
http://creativecommons.org/licenses/by/4.0/
We study data poisoning attacks in learning from human preferences. More specifically, we consider the problem of teaching/enforcing a target policy $\pi^\dagger$ by synthesizing preference data. We seek to understand the susceptibility of different preference-based learning paradigms to poisoned preference data by analyzing the number of samples required by the attacker to enforce $\pi^\dagger$. We first propose a general data poisoning formulation in learning from human preferences and then study it for two popular paradigms, namely: (a) reinforcement learning from human feedback (RLHF) that operates by learning a reward model using preferences; (b) direct preference optimization (DPO) that directly optimizes policy using preferences. We conduct a theoretical analysis of the effectiveness of data poisoning in a setting where the attacker is allowed to augment a pre-existing dataset and also study its special case where the attacker can synthesize the entire preference dataset from scratch. As our main results, we provide lower/upper bounds on the number of samples required to enforce $\pi^\dagger$. Finally, we discuss the implications of our results in terms of the susceptibility of these learning paradigms under such data poisoning attacks.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 10:11:54 GMT" } ]
2025-03-14T00:00:00
[ [ "Nika", "Andi", "" ], [ "Nöther", "Jonathan", "" ], [ "Mandal", "Debmalya", "" ], [ "Kamalaruban", "Parameswaran", "" ], [ "Singla", "Adish", "" ], [ "Radanović", "Goran", "" ] ]
TITLE: Policy Teaching via Data Poisoning in Learning from Human Preferences ABSTRACT: We study data poisoning attacks in learning from human preferences. More specifically, we consider the problem of teaching/enforcing a target policy $\pi^\dagger$ by synthesizing preference data. We seek to understand the susceptibility of different preference-based learning paradigms to poisoned preference data by analyzing the number of samples required by the attacker to enforce $\pi^\dagger$. We first propose a general data poisoning formulation in learning from human preferences and then study it for two popular paradigms, namely: (a) reinforcement learning from human feedback (RLHF) that operates by learning a reward model using preferences; (b) direct preference optimization (DPO) that directly optimizes policy using preferences. We conduct a theoretical analysis of the effectiveness of data poisoning in a setting where the attacker is allowed to augment a pre-existing dataset and also study its special case where the attacker can synthesize the entire preference dataset from scratch. As our main results, we provide lower/upper bounds on the number of samples required to enforce $\pi^\dagger$. Finally, we discuss the implications of our results in terms of the susceptibility of these learning paradigms under such data poisoning attacks.
2503.10233
Laya Mahmoudi
Samira Zangooei, Amirhossein Darmani, Hossein Farahmand Nezhad, Laya Mahmoudi
ARLED: Leveraging LED-based ARMAN Model for Abstractive Summarization of Persian Long Documents
11 pages, 3 tables
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The increasing volume of textual data poses challenges in reading and comprehending large documents, particularly for scholars who need to extract useful information from research articles. Automatic text summarization has emerged as a powerful tool to condense lengthy documents into concise and informative summaries. Depending on the approach used, text summarization can be categorized as either extractive or abstractive. While extractive methods are commonly used due to their simplicity, they often miss important information. On the other hand, Abstractive Summarization can generate more coherent and informative summaries by understanding the underlying meaning of the text. Abstractive techniques have gained attention in various languages, and recent advancements have been achieved through pre-training models such as BERT, BART, and T5. However, the challenge of summarizing long documents remains, and alternative models like Longformer have been introduced to address this limitation. In this context, this paper focuses on abstractive summarization in the Persian language. The authors introduce a new dataset of 300,000 full-text Persian papers obtained from the Ensani website and apply the ARMAN model, based on the Longformer architecture, to generate summaries. The experimental results demonstrate promising performance in Persian text summarization. The paper provides a comprehensive overview of related work, discusses the methodology, presents the experimental results, and concludes with future research directions.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 10:16:46 GMT" } ]
2025-03-14T00:00:00
[ [ "Zangooei", "Samira", "" ], [ "Darmani", "Amirhossein", "" ], [ "Nezhad", "Hossein Farahmand", "" ], [ "Mahmoudi", "Laya", "" ] ]
TITLE: ARLED: Leveraging LED-based ARMAN Model for Abstractive Summarization of Persian Long Documents ABSTRACT: The increasing volume of textual data poses challenges in reading and comprehending large documents, particularly for scholars who need to extract useful information from research articles. Automatic text summarization has emerged as a powerful tool to condense lengthy documents into concise and informative summaries. Depending on the approach used, text summarization can be categorized as either extractive or abstractive. While extractive methods are commonly used due to their simplicity, they often miss important information. On the other hand, Abstractive Summarization can generate more coherent and informative summaries by understanding the underlying meaning of the text. Abstractive techniques have gained attention in various languages, and recent advancements have been achieved through pre-training models such as BERT, BART, and T5. However, the challenge of summarizing long documents remains, and alternative models like Longformer have been introduced to address this limitation. In this context, this paper focuses on abstractive summarization in the Persian language. The authors introduce a new dataset of 300,000 full-text Persian papers obtained from the Ensani website and apply the ARMAN model, based on the Longformer architecture, to generate summaries. The experimental results demonstrate promising performance in Persian text summarization. The paper provides a comprehensive overview of related work, discusses the methodology, presents the experimental results, and concludes with future research directions.
2503.10239
Yifeng Cai
Yifeng Cai, Ziqi Zhang, Mengyu Yao, Junlin Liu, Xiaoke Zhao, Xinyi Fu, Ruoyu Li, Zhe Li, Xiangqun Chen, Yao Guo, Ding Li
I Can Tell Your Secrets: Inferring Privacy Attributes from Mini-app Interaction History in Super-apps
Accepted by USENIX Security 2025
null
null
null
cs.CR
http://creativecommons.org/licenses/by/4.0/
Super-apps have emerged as comprehensive platforms integrating various mini-apps to provide diverse services. While super-apps offer convenience and enriched functionality, they can introduce new privacy risks. This paper reveals a new privacy leakage source in super-apps: mini-app interaction history, including mini-app usage history (Mini-H) and operation history (Op-H). Mini-H refers to the history of mini-apps accessed by users, such as their frequency and categories. Op-H captures user interactions within mini-apps, including button clicks, bar drags, and image views. Super-apps can naturally collect these data without instrumentation due to the web-based feature of mini-apps. We identify these data types as novel and unexplored privacy risks through a literature review of 30 papers and an empirical analysis of 31 super-apps. We design a mini-app interaction history-oriented inference attack (THEFT), to exploit this new vulnerability. Using THEFT, the insider threats within the low-privilege business department of the super-app vendor acting as the adversary can achieve more than 95.5% accuracy in inferring privacy attributes of over 16.1% of users. THEFT only requires a small training dataset of 200 users from public breached databases on the Internet. We also engage with super-app vendors and a standards association to increase industry awareness and commitment to protect this data. Our contributions are significant in identifying overlooked privacy risks, demonstrating the effectiveness of a new attack, and influencing industry practices toward better privacy protection in the super-app ecosystem.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 10:29:40 GMT" } ]
2025-03-14T00:00:00
[ [ "Cai", "Yifeng", "" ], [ "Zhang", "Ziqi", "" ], [ "Yao", "Mengyu", "" ], [ "Liu", "Junlin", "" ], [ "Zhao", "Xiaoke", "" ], [ "Fu", "Xinyi", "" ], [ "Li", "Ruoyu", "" ], [ "Li", "Zhe", "" ...
TITLE: I Can Tell Your Secrets: Inferring Privacy Attributes from Mini-app Interaction History in Super-apps ABSTRACT: Super-apps have emerged as comprehensive platforms integrating various mini-apps to provide diverse services. While super-apps offer convenience and enriched functionality, they can introduce new privacy risks. This paper reveals a new privacy leakage source in super-apps: mini-app interaction history, including mini-app usage history (Mini-H) and operation history (Op-H). Mini-H refers to the history of mini-apps accessed by users, such as their frequency and categories. Op-H captures user interactions within mini-apps, including button clicks, bar drags, and image views. Super-apps can naturally collect these data without instrumentation due to the web-based feature of mini-apps. We identify these data types as novel and unexplored privacy risks through a literature review of 30 papers and an empirical analysis of 31 super-apps. We design a mini-app interaction history-oriented inference attack (THEFT), to exploit this new vulnerability. Using THEFT, the insider threats within the low-privilege business department of the super-app vendor acting as the adversary can achieve more than 95.5% accuracy in inferring privacy attributes of over 16.1% of users. THEFT only requires a small training dataset of 200 users from public breached databases on the Internet. We also engage with super-app vendors and a standards association to increase industry awareness and commitment to protect this data. Our contributions are significant in identifying overlooked privacy risks, demonstrating the effectiveness of a new attack, and influencing industry practices toward better privacy protection in the super-app ecosystem.
2503.10240
Bogdan Chornomaz
Bogdan Chornomaz, Shay Moran, Tom Waknine
Spherical dimension
null
null
null
null
cs.DM cs.LG
http://creativecommons.org/licenses/by-nc-sa/4.0/
We introduce and study the spherical dimension, a natural topological relaxation of the VC dimension that unifies several results in learning theory where topology plays a key role in the proofs. The spherical dimension is defined by extending the set of realizable datasets (used to define the VC dimension) to the continuous space of realizable distributions. In this space, a shattered set of size d (in the VC sense) is completed into a continuous object, specifically a d-dimensional sphere of realizable distributions. The spherical dimension is then defined as the dimension of the largest sphere in this space. Thus, the spherical dimension is at least the VC dimension. The spherical dimension serves as a common foundation for leveraging the Borsuk-Ulam theorem and related topological tools. We demonstrate the utility of the spherical dimension in diverse applications, including disambiguations of partial concept classes, reductions from classification to stochastic convex optimization, stability and replicability, and sample compression schemes. Perhaps surprisingly, we show that the open question posed by Alon, Hanneke, Holzman, and Moran (FOCS 2021) of whether there exist non-trivial disambiguations for halfspaces with margin is equivalent to the basic open question of whether the VC and spherical dimensions are finite together.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 10:32:25 GMT" } ]
2025-03-14T00:00:00
[ [ "Chornomaz", "Bogdan", "" ], [ "Moran", "Shay", "" ], [ "Waknine", "Tom", "" ] ]
TITLE: Spherical dimension ABSTRACT: We introduce and study the spherical dimension, a natural topological relaxation of the VC dimension that unifies several results in learning theory where topology plays a key role in the proofs. The spherical dimension is defined by extending the set of realizable datasets (used to define the VC dimension) to the continuous space of realizable distributions. In this space, a shattered set of size d (in the VC sense) is completed into a continuous object, specifically a d-dimensional sphere of realizable distributions. The spherical dimension is then defined as the dimension of the largest sphere in this space. Thus, the spherical dimension is at least the VC dimension. The spherical dimension serves as a common foundation for leveraging the Borsuk-Ulam theorem and related topological tools. We demonstrate the utility of the spherical dimension in diverse applications, including disambiguations of partial concept classes, reductions from classification to stochastic convex optimization, stability and replicability, and sample compression schemes. Perhaps surprisingly, we show that the open question posed by Alon, Hanneke, Holzman, and Moran (FOCS 2021) of whether there exist non-trivial disambiguations for halfspaces with margin is equivalent to the basic open question of whether the VC and spherical dimensions are finite together.
2503.10247
Zhijie Zhu
Zhijie Zhu, Lei Fan, Maurice Pagnucco, Yang Song
Interpretable Image Classification via Non-parametric Part Prototype Learning
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
Classifying images with an interpretable decision-making process is a long-standing problem in computer vision. In recent years, Prototypical Part Networks has gained traction as an approach for self-explainable neural networks, due to their ability to mimic human visual reasoning by providing explanations based on prototypical object parts. However, the quality of the explanations generated by these methods leaves room for improvement, as the prototypes usually focus on repetitive and redundant concepts. Leveraging recent advances in prototype learning, we present a framework for part-based interpretable image classification that learns a set of semantically distinctive object parts for each class, and provides diverse and comprehensive explanations. The core of our method is to learn the part-prototypes in a non-parametric fashion, through clustering deep features extracted from foundation vision models that encode robust semantic information. To quantitatively evaluate the quality of explanations provided by ProtoPNets, we introduce Distinctiveness Score and Comprehensiveness Score. Through evaluation on CUB-200-2011, Stanford Cars and Stanford Dogs datasets, we show that our framework compares favourably against existing ProtoPNets while achieving better interpretability. Code is available at: https://github.com/zijizhu/proto-non-param.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 10:46:53 GMT" } ]
2025-03-14T00:00:00
[ [ "Zhu", "Zhijie", "" ], [ "Fan", "Lei", "" ], [ "Pagnucco", "Maurice", "" ], [ "Song", "Yang", "" ] ]
TITLE: Interpretable Image Classification via Non-parametric Part Prototype Learning ABSTRACT: Classifying images with an interpretable decision-making process is a long-standing problem in computer vision. In recent years, Prototypical Part Networks has gained traction as an approach for self-explainable neural networks, due to their ability to mimic human visual reasoning by providing explanations based on prototypical object parts. However, the quality of the explanations generated by these methods leaves room for improvement, as the prototypes usually focus on repetitive and redundant concepts. Leveraging recent advances in prototype learning, we present a framework for part-based interpretable image classification that learns a set of semantically distinctive object parts for each class, and provides diverse and comprehensive explanations. The core of our method is to learn the part-prototypes in a non-parametric fashion, through clustering deep features extracted from foundation vision models that encode robust semantic information. To quantitatively evaluate the quality of explanations provided by ProtoPNets, we introduce Distinctiveness Score and Comprehensiveness Score. Through evaluation on CUB-200-2011, Stanford Cars and Stanford Dogs datasets, we show that our framework compares favourably against existing ProtoPNets while achieving better interpretability. Code is available at: https://github.com/zijizhu/proto-non-param.
2503.10256
Yeonjin Chang
Yeonjin Chang, Erqun Dong, Seunghyeon Seo, Nojun Kwak, Kwang Moo Yi
ROODI: Reconstructing Occluded Objects with Denoising Inpainters
Project page: https://yeonjin-chang.github.io/ROODI/
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
While the quality of novel-view images has improved dramatically with 3D Gaussian Splatting, extracting specific objects from scenes remains challenging. Isolating individual 3D Gaussian primitives for each object and handling occlusions in scenes remain far from being solved. We propose a novel object extraction method based on two key principles: (1) being object-centric by pruning irrelevant primitives; and (2) leveraging generative inpainting to compensate for missing observations caused by occlusions. For pruning, we analyze the local structure of primitives using K-nearest neighbors, and retain only relevant ones. For inpainting, we employ an off-the-shelf diffusion-based inpainter combined with occlusion reasoning, utilizing the 3D representation of the entire scene. Our findings highlight the crucial synergy between pruning and inpainting, both of which significantly enhance extraction performance. We evaluate our method on a standard real-world dataset and introduce a synthetic dataset for quantitative analysis. Our approach outperforms the state-of-the-art, demonstrating its effectiveness in object extraction from complex scenes.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 11:16:21 GMT" } ]
2025-03-14T00:00:00
[ [ "Chang", "Yeonjin", "" ], [ "Dong", "Erqun", "" ], [ "Seo", "Seunghyeon", "" ], [ "Kwak", "Nojun", "" ], [ "Yi", "Kwang Moo", "" ] ]
TITLE: ROODI: Reconstructing Occluded Objects with Denoising Inpainters ABSTRACT: While the quality of novel-view images has improved dramatically with 3D Gaussian Splatting, extracting specific objects from scenes remains challenging. Isolating individual 3D Gaussian primitives for each object and handling occlusions in scenes remain far from being solved. We propose a novel object extraction method based on two key principles: (1) being object-centric by pruning irrelevant primitives; and (2) leveraging generative inpainting to compensate for missing observations caused by occlusions. For pruning, we analyze the local structure of primitives using K-nearest neighbors, and retain only relevant ones. For inpainting, we employ an off-the-shelf diffusion-based inpainter combined with occlusion reasoning, utilizing the 3D representation of the entire scene. Our findings highlight the crucial synergy between pruning and inpainting, both of which significantly enhance extraction performance. We evaluate our method on a standard real-world dataset and introduce a synthetic dataset for quantitative analysis. Our approach outperforms the state-of-the-art, demonstrating its effectiveness in object extraction from complex scenes.
2503.10257
Zeyi Xu
Zeyi Xu, Jinfan Liu, Kuangxu Chen, Ye Chen, Zhangli Hu, Bingbing Ni
AMR-Transformer: Enabling Efficient Long-range Interaction for Complex Neural Fluid Simulation
null
null
null
null
cs.LG
http://creativecommons.org/licenses/by/4.0/
Accurately and efficiently simulating complex fluid dynamics is a challenging task that has traditionally relied on computationally intensive methods. Neural network-based approaches, such as convolutional and graph neural networks, have partially alleviated this burden by enabling efficient local feature extraction. However, they struggle to capture long-range dependencies due to limited receptive fields, and Transformer-based models, while providing global context, incur prohibitive computational costs. To tackle these challenges, we propose AMR-Transformer, an efficient and accurate neural CFD-solving pipeline that integrates a novel adaptive mesh refinement scheme with a Navier-Stokes constraint-aware fast pruning module. This design encourages long-range interactions between simulation cells and facilitates the modeling of global fluid wave patterns, such as turbulence and shockwaves. Experiments show that our approach achieves significant gains in efficiency while preserving critical details, making it suitable for high-resolution physical simulations with long-range dependencies. On CFDBench, PDEBench and a new shockwave dataset, our pipeline demonstrates up to an order-of-magnitude improvement in accuracy over baseline models. Additionally, compared to ViT, our approach achieves a reduction in FLOPs of up to 60 times.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 11:16:42 GMT" } ]
2025-03-14T00:00:00
[ [ "Xu", "Zeyi", "" ], [ "Liu", "Jinfan", "" ], [ "Chen", "Kuangxu", "" ], [ "Chen", "Ye", "" ], [ "Hu", "Zhangli", "" ], [ "Ni", "Bingbing", "" ] ]
TITLE: AMR-Transformer: Enabling Efficient Long-range Interaction for Complex Neural Fluid Simulation ABSTRACT: Accurately and efficiently simulating complex fluid dynamics is a challenging task that has traditionally relied on computationally intensive methods. Neural network-based approaches, such as convolutional and graph neural networks, have partially alleviated this burden by enabling efficient local feature extraction. However, they struggle to capture long-range dependencies due to limited receptive fields, and Transformer-based models, while providing global context, incur prohibitive computational costs. To tackle these challenges, we propose AMR-Transformer, an efficient and accurate neural CFD-solving pipeline that integrates a novel adaptive mesh refinement scheme with a Navier-Stokes constraint-aware fast pruning module. This design encourages long-range interactions between simulation cells and facilitates the modeling of global fluid wave patterns, such as turbulence and shockwaves. Experiments show that our approach achieves significant gains in efficiency while preserving critical details, making it suitable for high-resolution physical simulations with long-range dependencies. On CFDBench, PDEBench and a new shockwave dataset, our pipeline demonstrates up to an order-of-magnitude improvement in accuracy over baseline models. Additionally, compared to ViT, our approach achieves a reduction in FLOPs of up to 60 times.
2503.10259
Yunpeng Qu
Yunpeng Qu, Kun Yuan, Qizhi Xie, Ming Sun, Chao Zhou, Jian Wang
KVQ: Boosting Video Quality Assessment via Saliency-guided Local Perception
11 pages, 7 figures
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Video Quality Assessment (VQA), which intends to predict the perceptual quality of videos, has attracted increasing attention. Due to factors like motion blur or specific distortions, the quality of different regions in a video varies. Recognizing the region-wise local quality within a video is beneficial for assessing global quality and can guide us in adopting fine-grained enhancement or transcoding strategies. Due to the heavy cost of annotating region-wise quality, the lack of ground truth constraints from relevant datasets further complicates the utilization of local perception. Inspired by the Human Visual System (HVS) that links global quality to the local texture of different regions and their visual saliency, we propose a Kaleidoscope Video Quality Assessment (KVQ) framework, which aims to effectively assess both saliency and local texture, thereby facilitating the assessment of global quality. Our framework extracts visual saliency and allocates attention using Fusion-Window Attention (FWA) while incorporating a Local Perception Constraint (LPC) to mitigate the reliance of regional texture perception on neighboring areas. KVQ obtains significant improvements across multiple scenarios on five VQA benchmarks compared to SOTA methods. Furthermore, to assess local perception, we establish a new Local Perception Visual Quality (LPVQ) dataset with region-wise annotations. Experimental results demonstrate the capability of KVQ in perceiving local distortions. KVQ models and the LPVQ dataset will be available at https://github.com/qyp2000/KVQ.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 11:16:58 GMT" } ]
2025-03-14T00:00:00
[ [ "Qu", "Yunpeng", "" ], [ "Yuan", "Kun", "" ], [ "Xie", "Qizhi", "" ], [ "Sun", "Ming", "" ], [ "Zhou", "Chao", "" ], [ "Wang", "Jian", "" ] ]
TITLE: KVQ: Boosting Video Quality Assessment via Saliency-guided Local Perception ABSTRACT: Video Quality Assessment (VQA), which intends to predict the perceptual quality of videos, has attracted increasing attention. Due to factors like motion blur or specific distortions, the quality of different regions in a video varies. Recognizing the region-wise local quality within a video is beneficial for assessing global quality and can guide us in adopting fine-grained enhancement or transcoding strategies. Due to the heavy cost of annotating region-wise quality, the lack of ground truth constraints from relevant datasets further complicates the utilization of local perception. Inspired by the Human Visual System (HVS) that links global quality to the local texture of different regions and their visual saliency, we propose a Kaleidoscope Video Quality Assessment (KVQ) framework, which aims to effectively assess both saliency and local texture, thereby facilitating the assessment of global quality. Our framework extracts visual saliency and allocates attention using Fusion-Window Attention (FWA) while incorporating a Local Perception Constraint (LPC) to mitigate the reliance of regional texture perception on neighboring areas. KVQ obtains significant improvements across multiple scenarios on five VQA benchmarks compared to SOTA methods. Furthermore, to assess local perception, we establish a new Local Perception Visual Quality (LPVQ) dataset with region-wise annotations. Experimental results demonstrate the capability of KVQ in perceiving local distortions. KVQ models and the LPVQ dataset will be available at https://github.com/qyp2000/KVQ.
2503.10265
Chang Han Low
Chang Han Low, Ziyue Wang, Tianyi Zhang, Zhitao Zeng, Zhu Zhuo, Evangelos B. Mazomenos, Yueming Jin
SurgRAW: Multi-Agent Workflow with Chain-of-Thought Reasoning for Surgical Intelligence
null
null
null
null
cs.AI cs.RO
http://creativecommons.org/licenses/by/4.0/
Integration of Vision-Language Models (VLMs) in surgical intelligence is hindered by hallucinations, domain knowledge gaps, and limited understanding of task interdependencies within surgical scenes, undermining clinical reliability. While recent VLMs demonstrate strong general reasoning and thinking capabilities, they still lack the domain expertise and task-awareness required for precise surgical scene interpretation. Although Chain-of-Thought (CoT) can structure reasoning more effectively, current approaches rely on self-generated CoT steps, which often exacerbate inherent domain gaps and hallucinations. To overcome this, we present SurgRAW, a CoT-driven multi-agent framework that delivers transparent, interpretable insights for most tasks in robotic-assisted surgery. By employing specialized CoT prompts across five tasks: instrument recognition, action recognition, action prediction, patient data extraction, and outcome assessment, SurgRAW mitigates hallucinations through structured, domain-aware reasoning. Retrieval-Augmented Generation (RAG) is also integrated to external medical knowledge to bridge domain gaps and improve response reliability. Most importantly, a hierarchical agentic system ensures that CoT-embedded VLM agents collaborate effectively while understanding task interdependencies, with a panel discussion mechanism promotes logical consistency. To evaluate our method, we introduce SurgCoTBench, the first reasoning-based dataset with structured frame-level annotations. With comprehensive experiments, we demonstrate the effectiveness of proposed SurgRAW with 29.32% accuracy improvement over baseline VLMs on 12 robotic procedures, achieving the state-of-the-art performance and advancing explainable, trustworthy, and autonomous surgical assistance.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 11:23:13 GMT" } ]
2025-03-14T00:00:00
[ [ "Low", "Chang Han", "" ], [ "Wang", "Ziyue", "" ], [ "Zhang", "Tianyi", "" ], [ "Zeng", "Zhitao", "" ], [ "Zhuo", "Zhu", "" ], [ "Mazomenos", "Evangelos B.", "" ], [ "Jin", "Yueming", "" ] ]
TITLE: SurgRAW: Multi-Agent Workflow with Chain-of-Thought Reasoning for Surgical Intelligence ABSTRACT: Integration of Vision-Language Models (VLMs) in surgical intelligence is hindered by hallucinations, domain knowledge gaps, and limited understanding of task interdependencies within surgical scenes, undermining clinical reliability. While recent VLMs demonstrate strong general reasoning and thinking capabilities, they still lack the domain expertise and task-awareness required for precise surgical scene interpretation. Although Chain-of-Thought (CoT) can structure reasoning more effectively, current approaches rely on self-generated CoT steps, which often exacerbate inherent domain gaps and hallucinations. To overcome this, we present SurgRAW, a CoT-driven multi-agent framework that delivers transparent, interpretable insights for most tasks in robotic-assisted surgery. By employing specialized CoT prompts across five tasks: instrument recognition, action recognition, action prediction, patient data extraction, and outcome assessment, SurgRAW mitigates hallucinations through structured, domain-aware reasoning. Retrieval-Augmented Generation (RAG) is also integrated to external medical knowledge to bridge domain gaps and improve response reliability. Most importantly, a hierarchical agentic system ensures that CoT-embedded VLM agents collaborate effectively while understanding task interdependencies, with a panel discussion mechanism promotes logical consistency. To evaluate our method, we introduce SurgCoTBench, the first reasoning-based dataset with structured frame-level annotations. With comprehensive experiments, we demonstrate the effectiveness of proposed SurgRAW with 29.32% accuracy improvement over baseline VLMs on 12 robotic procedures, achieving the state-of-the-art performance and advancing explainable, trustworthy, and autonomous surgical assistance.
2503.10269
Wassim Bouaziz
Wassim Bouaziz, El-Mahdi El-Mhamdi, Nicolas Usunier
Targeted Data Poisoning for Black-Box Audio Datasets Ownership Verification
Published at ICASSP 2025, 5 pages, 7 figures
null
null
null
cs.CR cs.LG
http://creativecommons.org/licenses/by/4.0/
Protecting the use of audio datasets is a major concern for data owners, particularly with the recent rise of audio deep learning models. While watermarks can be used to protect the data itself, they do not allow to identify a deep learning model trained on a protected dataset. In this paper, we adapt to audio data the recently introduced data taggants approach. Data taggants is a method to verify if a neural network was trained on a protected image dataset with top-$k$ predictions access to the model only. This method relies on a targeted data poisoning scheme by discreetly altering a small fraction (1%) of the dataset as to induce a harmless behavior on out-of-distribution data called keys. We evaluate our method on the Speechcommands and the ESC50 datasets and state of the art transformer models, and show that we can detect the use of the dataset with high confidence without loss of performance. We also show the robustness of our method against common data augmentation techniques, making it a practical method to protect audio datasets.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 11:25:25 GMT" } ]
2025-03-14T00:00:00
[ [ "Bouaziz", "Wassim", "" ], [ "El-Mhamdi", "El-Mahdi", "" ], [ "Usunier", "Nicolas", "" ] ]
TITLE: Targeted Data Poisoning for Black-Box Audio Datasets Ownership Verification ABSTRACT: Protecting the use of audio datasets is a major concern for data owners, particularly with the recent rise of audio deep learning models. While watermarks can be used to protect the data itself, they do not allow to identify a deep learning model trained on a protected dataset. In this paper, we adapt to audio data the recently introduced data taggants approach. Data taggants is a method to verify if a neural network was trained on a protected image dataset with top-$k$ predictions access to the model only. This method relies on a targeted data poisoning scheme by discreetly altering a small fraction (1%) of the dataset as to induce a harmless behavior on out-of-distribution data called keys. We evaluate our method on the Speechcommands and the ESC50 datasets and state of the art transformer models, and show that we can detect the use of the dataset with high confidence without loss of performance. We also show the robustness of our method against common data augmentation techniques, making it a practical method to protect audio datasets.
2503.10284
Zhen Zhang
Zhen Zhang, Meihan Liu, Bingsheng He
PyGDA: A Python Library for Graph Domain Adaptation
Under Review
null
null
null
cs.LG cs.AI
http://creativecommons.org/licenses/by/4.0/
Graph domain adaptation has emerged as a promising approach to facilitate knowledge transfer across different domains. Recently, numerous models have been proposed to enhance their generalization capabilities in this field. However, there is still no unified library that brings together existing techniques and simplifies their implementation. To fill this gap, we introduce PyGDA, an open-source Python library tailored for graph domain adaptation. As the first comprehensive library in this area, PyGDA covers more than 20 widely used graph domain adaptation methods together with different types of graph datasets. Specifically, PyGDA offers modular components, enabling users to seamlessly build custom models with a variety of commonly used utility functions. To handle large-scale graphs, PyGDA includes support for features such as sampling and mini-batch processing, ensuring efficient computation. In addition, PyGDA also includes comprehensive performance benchmarks and well-documented user-friendly API for both researchers and practitioners. To foster convenient accessibility, PyGDA is released under the MIT license at https://github.com/pygda-team/pygda, and the API documentation is https://pygda.readthedocs.io/en/stable/.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 11:52:23 GMT" } ]
2025-03-14T00:00:00
[ [ "Zhang", "Zhen", "" ], [ "Liu", "Meihan", "" ], [ "He", "Bingsheng", "" ] ]
TITLE: PyGDA: A Python Library for Graph Domain Adaptation ABSTRACT: Graph domain adaptation has emerged as a promising approach to facilitate knowledge transfer across different domains. Recently, numerous models have been proposed to enhance their generalization capabilities in this field. However, there is still no unified library that brings together existing techniques and simplifies their implementation. To fill this gap, we introduce PyGDA, an open-source Python library tailored for graph domain adaptation. As the first comprehensive library in this area, PyGDA covers more than 20 widely used graph domain adaptation methods together with different types of graph datasets. Specifically, PyGDA offers modular components, enabling users to seamlessly build custom models with a variety of commonly used utility functions. To handle large-scale graphs, PyGDA includes support for features such as sampling and mini-batch processing, ensuring efficient computation. In addition, PyGDA also includes comprehensive performance benchmarks and well-documented user-friendly API for both researchers and practitioners. To foster convenient accessibility, PyGDA is released under the MIT license at https://github.com/pygda-team/pygda, and the API documentation is https://pygda.readthedocs.io/en/stable/.
2503.10286
Zhiqi Li
Zhiqi Li, Chengrui Dong, Yiming Chen, Zhangchi Huang, Peidong Liu
VicaSplat: A Single Run is All You Need for 3D Gaussian Splatting and Camera Estimation from Unposed Video Frames
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present VicaSplat, a novel framework for joint 3D Gaussians reconstruction and camera pose estimation from a sequence of unposed video frames, which is a critical yet underexplored task in real-world 3D applications. The core of our method lies in a novel transformer-based network architecture. In particular, our model starts with an image encoder that maps each image to a list of visual tokens. All visual tokens are concatenated with additional inserted learnable camera tokens. The obtained tokens then fully communicate with each other within a tailored transformer decoder. The camera tokens causally aggregate features from visual tokens of different views, and further modulate them frame-wisely to inject view-dependent features. 3D Gaussian splats and camera pose parameters can then be estimated via different prediction heads. Experiments show that VicaSplat surpasses baseline methods for multi-view inputs, and achieves comparable performance to prior two-view approaches. Remarkably, VicaSplat also demonstrates exceptional cross-dataset generalization capability on the ScanNet benchmark, achieving superior performance without any fine-tuning. Project page: https://lizhiqi49.github.io/VicaSplat.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 11:56:05 GMT" } ]
2025-03-14T00:00:00
[ [ "Li", "Zhiqi", "" ], [ "Dong", "Chengrui", "" ], [ "Chen", "Yiming", "" ], [ "Huang", "Zhangchi", "" ], [ "Liu", "Peidong", "" ] ]
TITLE: VicaSplat: A Single Run is All You Need for 3D Gaussian Splatting and Camera Estimation from Unposed Video Frames ABSTRACT: We present VicaSplat, a novel framework for joint 3D Gaussians reconstruction and camera pose estimation from a sequence of unposed video frames, which is a critical yet underexplored task in real-world 3D applications. The core of our method lies in a novel transformer-based network architecture. In particular, our model starts with an image encoder that maps each image to a list of visual tokens. All visual tokens are concatenated with additional inserted learnable camera tokens. The obtained tokens then fully communicate with each other within a tailored transformer decoder. The camera tokens causally aggregate features from visual tokens of different views, and further modulate them frame-wisely to inject view-dependent features. 3D Gaussian splats and camera pose parameters can then be estimated via different prediction heads. Experiments show that VicaSplat surpasses baseline methods for multi-view inputs, and achieves comparable performance to prior two-view approaches. Remarkably, VicaSplat also demonstrates exceptional cross-dataset generalization capability on the ScanNet benchmark, achieving superior performance without any fine-tuning. Project page: https://lizhiqi49.github.io/VicaSplat.
2503.10287
Hao Zhou
Hao Zhou, Xiaobao Guo, Yuzhe Zhu, Adams Wai-Kin Kong
MACS: Multi-source Audio-to-image Generation with Contextual Significance and Semantic Alignment
null
null
null
null
cs.SD cs.CV cs.GR eess.AS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Propelled by the breakthrough in deep generative models, audio-to-image generation has emerged as a pivotal cross-model task that converts complex auditory signals into rich visual representations. However, previous works only focus on single-source audio inputs for image generation, ignoring the multi-source characteristic in natural auditory scenes, thus limiting the performance in generating comprehensive visual content. To bridge this gap, a method called MACS is proposed to conduct multi-source audio-to-image generation. This is the first work that explicitly separates multi-source audio to capture the rich audio components before image generation. MACS is a two-stage method. In the first stage, multi-source audio inputs are separated by a weakly supervised method, where the audio and text labels are semantically aligned by casting into a common space using the large pre-trained CLAP model. We introduce a ranking loss to consider the contextual significance of the separated audio signals. In the second stage, efficient image generation is achieved by mapping the separated audio signals to the generation condition using only a trainable adapter and a MLP layer. We preprocess the LLP dataset as the first full multi-source audio-to-image generation benchmark. The experiments are conducted on multi-source, mixed-source, and single-source audio-to-image generation tasks. The proposed MACS outperforms the current state-of-the-art methods in 17 of the 21 evaluation indexes on all tasks and delivers superior visual quality. The code will be publicly available.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 11:56:25 GMT" } ]
2025-03-14T00:00:00
[ [ "Zhou", "Hao", "" ], [ "Guo", "Xiaobao", "" ], [ "Zhu", "Yuzhe", "" ], [ "Kong", "Adams Wai-Kin", "" ] ]
TITLE: MACS: Multi-source Audio-to-image Generation with Contextual Significance and Semantic Alignment ABSTRACT: Propelled by the breakthrough in deep generative models, audio-to-image generation has emerged as a pivotal cross-model task that converts complex auditory signals into rich visual representations. However, previous works only focus on single-source audio inputs for image generation, ignoring the multi-source characteristic in natural auditory scenes, thus limiting the performance in generating comprehensive visual content. To bridge this gap, a method called MACS is proposed to conduct multi-source audio-to-image generation. This is the first work that explicitly separates multi-source audio to capture the rich audio components before image generation. MACS is a two-stage method. In the first stage, multi-source audio inputs are separated by a weakly supervised method, where the audio and text labels are semantically aligned by casting into a common space using the large pre-trained CLAP model. We introduce a ranking loss to consider the contextual significance of the separated audio signals. In the second stage, efficient image generation is achieved by mapping the separated audio signals to the generation condition using only a trainable adapter and a MLP layer. We preprocess the LLP dataset as the first full multi-source audio-to-image generation benchmark. The experiments are conducted on multi-source, mixed-source, and single-source audio-to-image generation tasks. The proposed MACS outperforms the current state-of-the-art methods in 17 of the 21 evaluation indexes on all tasks and delivers superior visual quality. The code will be publicly available.
2503.10291
Weiyun Wang
Weiyun Wang, Zhangwei Gao, Lianjie Chen, Zhe Chen, Jinguo Zhu, Xiangyu Zhao, Yangzhou Liu, Yue Cao, Shenglong Ye, Xizhou Zhu, Lewei Lu, Haodong Duan, Yu Qiao, Jifeng Dai, Wenhai Wang
VisualPRM: An Effective Process Reward Model for Multimodal Reasoning
null
null
null
null
cs.CV cs.CL
http://creativecommons.org/licenses/by/4.0/
We introduce VisualPRM, an advanced multimodal Process Reward Model (PRM) with 8B parameters, which improves the reasoning abilities of existing Multimodal Large Language Models (MLLMs) across different model scales and families with Best-of-N (BoN) evaluation strategies. Specifically, our model improves the reasoning performance of three types of MLLMs and four different model scales. Even when applied to the highly capable InternVL2.5-78B, it achieves a 5.9-point improvement across seven multimodal reasoning benchmarks. Experimental results show that our model exhibits superior performance compared to Outcome Reward Models and Self-Consistency during BoN evaluation. To facilitate the training of multimodal PRMs, we construct a multimodal process supervision dataset VisualPRM400K using an automated data pipeline. For the evaluation of multimodal PRMs, we propose VisualProcessBench, a benchmark with human-annotated step-wise correctness labels, to measure the abilities of PRMs to detect erroneous steps in multimodal reasoning tasks. We hope that our work can inspire more future research and contribute to the development of MLLMs. Our model, data, and benchmark are released in https://internvl.github.io/blog/2025-03-13-VisualPRM/.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 12:03:37 GMT" } ]
2025-03-14T00:00:00
[ [ "Wang", "Weiyun", "" ], [ "Gao", "Zhangwei", "" ], [ "Chen", "Lianjie", "" ], [ "Chen", "Zhe", "" ], [ "Zhu", "Jinguo", "" ], [ "Zhao", "Xiangyu", "" ], [ "Liu", "Yangzhou", "" ], [ "Cao", "Yue"...
TITLE: VisualPRM: An Effective Process Reward Model for Multimodal Reasoning ABSTRACT: We introduce VisualPRM, an advanced multimodal Process Reward Model (PRM) with 8B parameters, which improves the reasoning abilities of existing Multimodal Large Language Models (MLLMs) across different model scales and families with Best-of-N (BoN) evaluation strategies. Specifically, our model improves the reasoning performance of three types of MLLMs and four different model scales. Even when applied to the highly capable InternVL2.5-78B, it achieves a 5.9-point improvement across seven multimodal reasoning benchmarks. Experimental results show that our model exhibits superior performance compared to Outcome Reward Models and Self-Consistency during BoN evaluation. To facilitate the training of multimodal PRMs, we construct a multimodal process supervision dataset VisualPRM400K using an automated data pipeline. For the evaluation of multimodal PRMs, we propose VisualProcessBench, a benchmark with human-annotated step-wise correctness labels, to measure the abilities of PRMs to detect erroneous steps in multimodal reasoning tasks. We hope that our work can inspire more future research and contribute to the development of MLLMs. Our model, data, and benchmark are released in https://internvl.github.io/blog/2025-03-13-VisualPRM/.
2503.10301
Moreno La Quatra
Moreno La Quatra, Juan Rafael Orozco-Arroyave, Marco Sabato Siniscalchi
Bilingual Dual-Head Deep Model for Parkinson's Disease Detection from Speech
Accepted at ICASSP 2025 - Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses
null
10.1109/ICASSP49660.2025.10889445
null
eess.AS cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This work aims to tackle the Parkinson's disease (PD) detection problem from the speech signal in a bilingual setting by proposing an ad-hoc dual-head deep neural architecture for type-based binary classification. One head is specialized for diadochokinetic patterns. The other head looks for natural speech patterns present in continuous spoken utterances. Only one of the two heads is operative accordingly to the nature of the input. Speech representations are extracted from self-supervised learning (SSL) models and wavelet transforms. Adaptive layers, convolutional bottlenecks, and contrastive learning are exploited to reduce variations across languages. Our solution is assessed against two distinct datasets, EWA-DB, and PC-GITA, which cover Slovak and Spanish languages, respectively. Results indicate that conventional models trained on a single language dataset struggle with cross-linguistic generalization, and naive combinations of datasets are suboptimal. In contrast, our model improves generalization on both languages, simultaneously.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 12:23:11 GMT" } ]
2025-03-14T00:00:00
[ [ "La Quatra", "Moreno", "" ], [ "Orozco-Arroyave", "Juan Rafael", "" ], [ "Siniscalchi", "Marco Sabato", "" ] ]
TITLE: Bilingual Dual-Head Deep Model for Parkinson's Disease Detection from Speech ABSTRACT: This work aims to tackle the Parkinson's disease (PD) detection problem from the speech signal in a bilingual setting by proposing an ad-hoc dual-head deep neural architecture for type-based binary classification. One head is specialized for diadochokinetic patterns. The other head looks for natural speech patterns present in continuous spoken utterances. Only one of the two heads is operative accordingly to the nature of the input. Speech representations are extracted from self-supervised learning (SSL) models and wavelet transforms. Adaptive layers, convolutional bottlenecks, and contrastive learning are exploited to reduce variations across languages. Our solution is assessed against two distinct datasets, EWA-DB, and PC-GITA, which cover Slovak and Spanish languages, respectively. Results indicate that conventional models trained on a single language dataset struggle with cross-linguistic generalization, and naive combinations of datasets are suboptimal. In contrast, our model improves generalization on both languages, simultaneously.
2503.10305
Emil Mededovic
Emil Mededovic, Yuli Wu, Henning Konermann, Marcin Kopaczka, Mareike Schulz, Rene Tolba, Johannes Stegmaier
Eye on the Target: Eye Tracking Meets Rodent Tracking
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Analyzing animal behavior from video recordings is crucial for scientific research, yet manual annotation remains labor-intensive and prone to subjectivity. Efficient segmentation methods are needed to automate this process while maintaining high accuracy. In this work, we propose a novel pipeline that utilizes eye-tracking data from Aria glasses to generate prompt points, which are then used to produce segmentation masks via a fast zero-shot segmentation model. Additionally, we apply post-processing to refine the prompts, leading to improved segmentation quality. Through our approach, we demonstrate that combining eye-tracking-based annotation with smart prompt refinement can enhance segmentation accuracy, achieving an improvement of 70.6% from 38.8 to 66.2 in the Jaccard Index for segmentation results in the rats dataset.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 12:27:42 GMT" } ]
2025-03-14T00:00:00
[ [ "Mededovic", "Emil", "" ], [ "Wu", "Yuli", "" ], [ "Konermann", "Henning", "" ], [ "Kopaczka", "Marcin", "" ], [ "Schulz", "Mareike", "" ], [ "Tolba", "Rene", "" ], [ "Stegmaier", "Johannes", "" ] ]
TITLE: Eye on the Target: Eye Tracking Meets Rodent Tracking ABSTRACT: Analyzing animal behavior from video recordings is crucial for scientific research, yet manual annotation remains labor-intensive and prone to subjectivity. Efficient segmentation methods are needed to automate this process while maintaining high accuracy. In this work, we propose a novel pipeline that utilizes eye-tracking data from Aria glasses to generate prompt points, which are then used to produce segmentation masks via a fast zero-shot segmentation model. Additionally, we apply post-processing to refine the prompts, leading to improved segmentation quality. Through our approach, we demonstrate that combining eye-tracking-based annotation with smart prompt refinement can enhance segmentation accuracy, achieving an improvement of 70.6% from 38.8 to 66.2 in the Jaccard Index for segmentation results in the rats dataset.
2503.10307
Martin C\'ifka
Georgy Ponimatkin, Martin C\'ifka, Tom\'a\v{s} Sou\v{c}ek, M\'ed\'eric Fourmy, Yann Labb\'e, Vladimir Petrik, Josef Sivic
6D Object Pose Tracking in Internet Videos for Robotic Manipulation
Accepted to ICLR 2025. Project page available at https://ponimatkin.github.io/wildpose/
null
null
null
cs.CV cs.RO
http://creativecommons.org/licenses/by/4.0/
We seek to extract a temporally consistent 6D pose trajectory of a manipulated object from an Internet instructional video. This is a challenging set-up for current 6D pose estimation methods due to uncontrolled capturing conditions, subtle but dynamic object motions, and the fact that the exact mesh of the manipulated object is not known. To address these challenges, we present the following contributions. First, we develop a new method that estimates the 6D pose of any object in the input image without prior knowledge of the object itself. The method proceeds by (i) retrieving a CAD model similar to the depicted object from a large-scale model database, (ii) 6D aligning the retrieved CAD model with the input image, and (iii) grounding the absolute scale of the object with respect to the scene. Second, we extract smooth 6D object trajectories from Internet videos by carefully tracking the detected objects across video frames. The extracted object trajectories are then retargeted via trajectory optimization into the configuration space of a robotic manipulator. Third, we thoroughly evaluate and ablate our 6D pose estimation method on YCB-V and HOPE-Video datasets as well as a new dataset of instructional videos manually annotated with approximate 6D object trajectories. We demonstrate significant improvements over existing state-of-the-art RGB 6D pose estimation methods. Finally, we show that the 6D object motion estimated from Internet videos can be transferred to a 7-axis robotic manipulator both in a virtual simulator as well as in a real world set-up. We also successfully apply our method to egocentric videos taken from the EPIC-KITCHENS dataset, demonstrating potential for Embodied AI applications.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 12:33:34 GMT" } ]
2025-03-14T00:00:00
[ [ "Ponimatkin", "Georgy", "" ], [ "Cífka", "Martin", "" ], [ "Souček", "Tomáš", "" ], [ "Fourmy", "Médéric", "" ], [ "Labbé", "Yann", "" ], [ "Petrik", "Vladimir", "" ], [ "Sivic", "Josef", "" ] ]
TITLE: 6D Object Pose Tracking in Internet Videos for Robotic Manipulation ABSTRACT: We seek to extract a temporally consistent 6D pose trajectory of a manipulated object from an Internet instructional video. This is a challenging set-up for current 6D pose estimation methods due to uncontrolled capturing conditions, subtle but dynamic object motions, and the fact that the exact mesh of the manipulated object is not known. To address these challenges, we present the following contributions. First, we develop a new method that estimates the 6D pose of any object in the input image without prior knowledge of the object itself. The method proceeds by (i) retrieving a CAD model similar to the depicted object from a large-scale model database, (ii) 6D aligning the retrieved CAD model with the input image, and (iii) grounding the absolute scale of the object with respect to the scene. Second, we extract smooth 6D object trajectories from Internet videos by carefully tracking the detected objects across video frames. The extracted object trajectories are then retargeted via trajectory optimization into the configuration space of a robotic manipulator. Third, we thoroughly evaluate and ablate our 6D pose estimation method on YCB-V and HOPE-Video datasets as well as a new dataset of instructional videos manually annotated with approximate 6D object trajectories. We demonstrate significant improvements over existing state-of-the-art RGB 6D pose estimation methods. Finally, we show that the 6D object motion estimated from Internet videos can be transferred to a 7-axis robotic manipulator both in a virtual simulator as well as in a real world set-up. We also successfully apply our method to egocentric videos taken from the EPIC-KITCHENS dataset, demonstrating potential for Embodied AI applications.
2503.10322
Haoxuan Li
Haoxuan Li, Sixu Yan, Yuhan Li, Xinggang Wang
Towards Fast, Memory-based and Data-Efficient Vision-Language Policy
11 pages, 7 figures, 6 tables
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Vision Language Models (VLMs) pretrained on Internet-scale vision-language data have demonstrated the potential to transfer their knowledge to robotic learning. However, the existing paradigm encounters three critical challenges: (1) expensive inference cost resulting from large-scale model parameters, (2) frequent domain shifts caused by mismatched data modalities, and (3) limited capacity to handle past or future experiences. In this work, we propose LiteVLP, a lightweight, memory-based, and general-purpose vision-language policy generation model. LiteVLP is built upon a pre-trained 1B-parameter VLM and fine-tuned on a tiny-scale and conversation-style robotic dataset. Through extensive experiments, we demonstrate that LiteVLP outperforms state-of-the-art vision-language policy on VIMA-Bench, with minimal training time. Furthermore, LiteVLP exhibits superior inference speed while maintaining exceptional high accuracy. In long-horizon manipulation tasks, LiteVLP also shows remarkable memory ability, outperforming the best-performing baseline model by 18.8%. These results highlight LiteVLP as a promising model to integrating the intelligence of VLMs into robotic learning.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 12:58:40 GMT" } ]
2025-03-14T00:00:00
[ [ "Li", "Haoxuan", "" ], [ "Yan", "Sixu", "" ], [ "Li", "Yuhan", "" ], [ "Wang", "Xinggang", "" ] ]
TITLE: Towards Fast, Memory-based and Data-Efficient Vision-Language Policy ABSTRACT: Vision Language Models (VLMs) pretrained on Internet-scale vision-language data have demonstrated the potential to transfer their knowledge to robotic learning. However, the existing paradigm encounters three critical challenges: (1) expensive inference cost resulting from large-scale model parameters, (2) frequent domain shifts caused by mismatched data modalities, and (3) limited capacity to handle past or future experiences. In this work, we propose LiteVLP, a lightweight, memory-based, and general-purpose vision-language policy generation model. LiteVLP is built upon a pre-trained 1B-parameter VLM and fine-tuned on a tiny-scale and conversation-style robotic dataset. Through extensive experiments, we demonstrate that LiteVLP outperforms state-of-the-art vision-language policy on VIMA-Bench, with minimal training time. Furthermore, LiteVLP exhibits superior inference speed while maintaining exceptional high accuracy. In long-horizon manipulation tasks, LiteVLP also shows remarkable memory ability, outperforming the best-performing baseline model by 18.8%. These results highlight LiteVLP as a promising model to integrating the intelligence of VLMs into robotic learning.
2503.10331
Maxim Popov
Maxim Popov, Regina Kurkova, Mikhail Iumanov, Jaafar Mahmoud, Sergey Kolyubin
OSMa-Bench: Evaluating Open Semantic Mapping Under Varying Lighting Conditions
Project page: https://be2rlab.github.io/OSMa-Bench/
null
null
null
cs.CV cs.AI cs.CL cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Open Semantic Mapping (OSM) is a key technology in robotic perception, combining semantic segmentation and SLAM techniques. This paper introduces a dynamically configurable and highly automated LLM/LVLM-powered pipeline for evaluating OSM solutions called OSMa-Bench (Open Semantic Mapping Benchmark). The study focuses on evaluating state-of-the-art semantic mapping algorithms under varying indoor lighting conditions, a critical challenge in indoor environments. We introduce a novel dataset with simulated RGB-D sequences and ground truth 3D reconstructions, facilitating the rigorous analysis of mapping performance across different lighting conditions. Through experiments on leading models such as ConceptGraphs, BBQ and OpenScene, we evaluate the semantic fidelity of object recognition and segmentation. Additionally, we introduce a Scene Graph evaluation method to analyze the ability of models to interpret semantic structure. The results provide insights into the robustness of these models, forming future research directions for developing resilient and adaptable robotic systems. Our code is available at https://be2rlab.github.io/OSMa-Bench/.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 13:07:51 GMT" } ]
2025-03-14T00:00:00
[ [ "Popov", "Maxim", "" ], [ "Kurkova", "Regina", "" ], [ "Iumanov", "Mikhail", "" ], [ "Mahmoud", "Jaafar", "" ], [ "Kolyubin", "Sergey", "" ] ]
TITLE: OSMa-Bench: Evaluating Open Semantic Mapping Under Varying Lighting Conditions ABSTRACT: Open Semantic Mapping (OSM) is a key technology in robotic perception, combining semantic segmentation and SLAM techniques. This paper introduces a dynamically configurable and highly automated LLM/LVLM-powered pipeline for evaluating OSM solutions called OSMa-Bench (Open Semantic Mapping Benchmark). The study focuses on evaluating state-of-the-art semantic mapping algorithms under varying indoor lighting conditions, a critical challenge in indoor environments. We introduce a novel dataset with simulated RGB-D sequences and ground truth 3D reconstructions, facilitating the rigorous analysis of mapping performance across different lighting conditions. Through experiments on leading models such as ConceptGraphs, BBQ and OpenScene, we evaluate the semantic fidelity of object recognition and segmentation. Additionally, we introduce a Scene Graph evaluation method to analyze the ability of models to interpret semantic structure. The results provide insights into the robustness of these models, forming future research directions for developing resilient and adaptable robotic systems. Our code is available at https://be2rlab.github.io/OSMa-Bench/.
2503.10350
Ali Salar
Ali Salar, Qing Liu, Yingli Tian and Guoying Zhao
Enhancing Facial Privacy Protection via Weakening Diffusion Purification
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The rapid growth of social media has led to the widespread sharing of individual portrait images, which pose serious privacy risks due to the capabilities of automatic face recognition (AFR) systems for mass surveillance. Hence, protecting facial privacy against unauthorized AFR systems is essential. Inspired by the generation capability of the emerging diffusion models, recent methods employ diffusion models to generate adversarial face images for privacy protection. However, they suffer from the diffusion purification effect, leading to a low protection success rate (PSR). In this paper, we first propose learning unconditional embeddings to increase the learning capacity for adversarial modifications and then use them to guide the modification of the adversarial latent code to weaken the diffusion purification effect. Moreover, we integrate an identity-preserving structure to maintain structural consistency between the original and generated images, allowing human observers to recognize the generated image as having the same identity as the original. Extensive experiments conducted on two public datasets, i.e., CelebA-HQ and LADN, demonstrate the superiority of our approach. The protected faces generated by our method outperform those produced by existing facial privacy protection approaches in terms of transferability and natural appearance.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 13:27:53 GMT" } ]
2025-03-14T00:00:00
[ [ "Salar", "Ali", "" ], [ "Liu", "Qing", "" ], [ "Tian", "Yingli", "" ], [ "Zhao", "Guoying", "" ] ]
TITLE: Enhancing Facial Privacy Protection via Weakening Diffusion Purification ABSTRACT: The rapid growth of social media has led to the widespread sharing of individual portrait images, which pose serious privacy risks due to the capabilities of automatic face recognition (AFR) systems for mass surveillance. Hence, protecting facial privacy against unauthorized AFR systems is essential. Inspired by the generation capability of the emerging diffusion models, recent methods employ diffusion models to generate adversarial face images for privacy protection. However, they suffer from the diffusion purification effect, leading to a low protection success rate (PSR). In this paper, we first propose learning unconditional embeddings to increase the learning capacity for adversarial modifications and then use them to guide the modification of the adversarial latent code to weaken the diffusion purification effect. Moreover, we integrate an identity-preserving structure to maintain structural consistency between the original and generated images, allowing human observers to recognize the generated image as having the same identity as the original. Extensive experiments conducted on two public datasets, i.e., CelebA-HQ and LADN, demonstrate the superiority of our approach. The protected faces generated by our method outperform those produced by existing facial privacy protection approaches in terms of transferability and natural appearance.
2503.10356
Toni Schneidereit
Toni Schneidereit, Stefan Gohrenz, Michael Breu{\ss}
Object detection characteristics in a learning factory environment using YOLOv8
null
null
null
null
cs.CV cs.AI
http://creativecommons.org/licenses/by/4.0/
AI-based object detection, and efforts to explain and investigate their characteristics, is a topic of high interest. The impact of, e.g., complex background structures with similar appearances as the objects of interest, on the detection accuracy and, beforehand, the necessary dataset composition are topics of ongoing research. In this paper, we present a systematic investigation of background influences and different features of the object to be detected. The latter includes various materials and surfaces, partially transparent and with shiny reflections in the context of an Industry 4.0 learning factory. Different YOLOv8 models have been trained for each of the materials on different sized datasets, where the appearance was the only changing parameter. In the end, similar characteristics tend to show different behaviours and sometimes unexpected results. While some background components tend to be detected, others with the same features are not part of the detection. Additionally, some more precise conclusions can be drawn from the results. Therefore, we contribute a challenging dataset with detailed investigations on 92 trained YOLO models, addressing some issues on the detection accuracy and possible overfitting.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 13:33:27 GMT" } ]
2025-03-14T00:00:00
[ [ "Schneidereit", "Toni", "" ], [ "Gohrenz", "Stefan", "" ], [ "Breuß", "Michael", "" ] ]
TITLE: Object detection characteristics in a learning factory environment using YOLOv8 ABSTRACT: AI-based object detection, and efforts to explain and investigate their characteristics, is a topic of high interest. The impact of, e.g., complex background structures with similar appearances as the objects of interest, on the detection accuracy and, beforehand, the necessary dataset composition are topics of ongoing research. In this paper, we present a systematic investigation of background influences and different features of the object to be detected. The latter includes various materials and surfaces, partially transparent and with shiny reflections in the context of an Industry 4.0 learning factory. Different YOLOv8 models have been trained for each of the materials on different sized datasets, where the appearance was the only changing parameter. In the end, similar characteristics tend to show different behaviours and sometimes unexpected results. While some background components tend to be detected, others with the same features are not part of the detection. Additionally, some more precise conclusions can be drawn from the results. Therefore, we contribute a challenging dataset with detailed investigations on 92 trained YOLO models, addressing some issues on the detection accuracy and possible overfitting.
2503.10370
Iman Nematollahi
Iman Nematollahi, Branton DeMoss, Akshay L Chandra, Nick Hawes, Wolfram Burgard, Ingmar Posner
LUMOS: Language-Conditioned Imitation Learning with World Models
Accepted at the 2025 IEEE International Conference on Robotics and Automation (ICRA)
null
null
null
cs.RO cs.CV cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We introduce LUMOS, a language-conditioned multi-task imitation learning framework for robotics. LUMOS learns skills by practicing them over many long-horizon rollouts in the latent space of a learned world model and transfers these skills zero-shot to a real robot. By learning on-policy in the latent space of the learned world model, our algorithm mitigates policy-induced distribution shift which most offline imitation learning methods suffer from. LUMOS learns from unstructured play data with fewer than 1% hindsight language annotations but is steerable with language commands at test time. We achieve this coherent long-horizon performance by combining latent planning with both image- and language-based hindsight goal relabeling during training, and by optimizing an intrinsic reward defined in the latent space of the world model over multiple time steps, effectively reducing covariate shift. In experiments on the difficult long-horizon CALVIN benchmark, LUMOS outperforms prior learning-based methods with comparable approaches on chained multi-task evaluations. To the best of our knowledge, we are the first to learn a language-conditioned continuous visuomotor control for a real-world robot within an offline world model. Videos, dataset and code are available at http://lumos.cs.uni-freiburg.de.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 13:48:24 GMT" } ]
2025-03-14T00:00:00
[ [ "Nematollahi", "Iman", "" ], [ "DeMoss", "Branton", "" ], [ "Chandra", "Akshay L", "" ], [ "Hawes", "Nick", "" ], [ "Burgard", "Wolfram", "" ], [ "Posner", "Ingmar", "" ] ]
TITLE: LUMOS: Language-Conditioned Imitation Learning with World Models ABSTRACT: We introduce LUMOS, a language-conditioned multi-task imitation learning framework for robotics. LUMOS learns skills by practicing them over many long-horizon rollouts in the latent space of a learned world model and transfers these skills zero-shot to a real robot. By learning on-policy in the latent space of the learned world model, our algorithm mitigates policy-induced distribution shift which most offline imitation learning methods suffer from. LUMOS learns from unstructured play data with fewer than 1% hindsight language annotations but is steerable with language commands at test time. We achieve this coherent long-horizon performance by combining latent planning with both image- and language-based hindsight goal relabeling during training, and by optimizing an intrinsic reward defined in the latent space of the world model over multiple time steps, effectively reducing covariate shift. In experiments on the difficult long-horizon CALVIN benchmark, LUMOS outperforms prior learning-based methods with comparable approaches on chained multi-task evaluations. To the best of our knowledge, we are the first to learn a language-conditioned continuous visuomotor control for a real-world robot within an offline world model. Videos, dataset and code are available at http://lumos.cs.uni-freiburg.de.
2503.10391
Yufan Deng
Yufan Deng, Xun Guo, Yizhi Wang, Jacob Zhiyuan Fang, Angtian Wang, Shenghai Yuan, Yiding Yang, Bo Liu, Haibin Huang, Chongyang Ma
CINEMA: Coherent Multi-Subject Video Generation via MLLM-Based Guidance
null
null
null
null
cs.CV cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Video generation has witnessed remarkable progress with the advent of deep generative models, particularly diffusion models. While existing methods excel in generating high-quality videos from text prompts or single images, personalized multi-subject video generation remains a largely unexplored challenge. This task involves synthesizing videos that incorporate multiple distinct subjects, each defined by separate reference images, while ensuring temporal and spatial consistency. Current approaches primarily rely on mapping subject images to keywords in text prompts, which introduces ambiguity and limits their ability to model subject relationships effectively. In this paper, we propose CINEMA, a novel framework for coherent multi-subject video generation by leveraging Multimodal Large Language Model (MLLM). Our approach eliminates the need for explicit correspondences between subject images and text entities, mitigating ambiguity and reducing annotation effort. By leveraging MLLM to interpret subject relationships, our method facilitates scalability, enabling the use of large and diverse datasets for training. Furthermore, our framework can be conditioned on varying numbers of subjects, offering greater flexibility in personalized content creation. Through extensive evaluations, we demonstrate that our approach significantly improves subject consistency, and overall video coherence, paving the way for advanced applications in storytelling, interactive media, and personalized video generation.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 14:07:58 GMT" } ]
2025-03-14T00:00:00
[ [ "Deng", "Yufan", "" ], [ "Guo", "Xun", "" ], [ "Wang", "Yizhi", "" ], [ "Fang", "Jacob Zhiyuan", "" ], [ "Wang", "Angtian", "" ], [ "Yuan", "Shenghai", "" ], [ "Yang", "Yiding", "" ], [ "Liu", "...
TITLE: CINEMA: Coherent Multi-Subject Video Generation via MLLM-Based Guidance ABSTRACT: Video generation has witnessed remarkable progress with the advent of deep generative models, particularly diffusion models. While existing methods excel in generating high-quality videos from text prompts or single images, personalized multi-subject video generation remains a largely unexplored challenge. This task involves synthesizing videos that incorporate multiple distinct subjects, each defined by separate reference images, while ensuring temporal and spatial consistency. Current approaches primarily rely on mapping subject images to keywords in text prompts, which introduces ambiguity and limits their ability to model subject relationships effectively. In this paper, we propose CINEMA, a novel framework for coherent multi-subject video generation by leveraging Multimodal Large Language Model (MLLM). Our approach eliminates the need for explicit correspondences between subject images and text entities, mitigating ambiguity and reducing annotation effort. By leveraging MLLM to interpret subject relationships, our method facilitates scalability, enabling the use of large and diverse datasets for training. Furthermore, our framework can be conditioned on varying numbers of subjects, offering greater flexibility in personalized content creation. Through extensive evaluations, we demonstrate that our approach significantly improves subject consistency, and overall video coherence, paving the way for advanced applications in storytelling, interactive media, and personalized video generation.
2503.10392
Fengxiang Wang
Fengxiang Wang, Hongzhen Wang, Yulin Wang, Di Wang, Mingshuo Chen, Haiyan Zhao, Yangang Sun, Shuo Wang, Long Lan, Wenjing Yang, Jing Zhang
RoMA: Scaling up Mamba-based Foundation Models for Remote Sensing
null
null
null
null
cs.CV cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Recent advances in self-supervised learning for Vision Transformers (ViTs) have fueled breakthroughs in remote sensing (RS) foundation models. However, the quadratic complexity of self-attention poses a significant barrier to scalability, particularly for large models and high-resolution images. While the linear-complexity Mamba architecture offers a promising alternative, existing RS applications of Mamba remain limited to supervised tasks on small, domain-specific datasets. To address these challenges, we propose RoMA, a framework that enables scalable self-supervised pretraining of Mamba-based RS foundation models using large-scale, diverse, unlabeled data. RoMA enhances scalability for high-resolution images through a tailored auto-regressive learning strategy, incorporating two key innovations: 1) a rotation-aware pretraining mechanism combining adaptive cropping with angular embeddings to handle sparsely distributed objects with arbitrary orientations, and 2) multi-scale token prediction objectives that address the extreme variations in object scales inherent to RS imagery. Systematic empirical studies validate that Mamba adheres to RS data and parameter scaling laws, with performance scaling reliably as model and data size increase. Furthermore, experiments across scene classification, object detection, and semantic segmentation tasks demonstrate that RoMA-pretrained Mamba models consistently outperform ViT-based counterparts in both accuracy and computational efficiency. The source code and pretrained models will be released at https://github.com/MiliLab/RoMA.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 14:09:18 GMT" } ]
2025-03-14T00:00:00
[ [ "Wang", "Fengxiang", "" ], [ "Wang", "Hongzhen", "" ], [ "Wang", "Yulin", "" ], [ "Wang", "Di", "" ], [ "Chen", "Mingshuo", "" ], [ "Zhao", "Haiyan", "" ], [ "Sun", "Yangang", "" ], [ "Wang", "S...
TITLE: RoMA: Scaling up Mamba-based Foundation Models for Remote Sensing ABSTRACT: Recent advances in self-supervised learning for Vision Transformers (ViTs) have fueled breakthroughs in remote sensing (RS) foundation models. However, the quadratic complexity of self-attention poses a significant barrier to scalability, particularly for large models and high-resolution images. While the linear-complexity Mamba architecture offers a promising alternative, existing RS applications of Mamba remain limited to supervised tasks on small, domain-specific datasets. To address these challenges, we propose RoMA, a framework that enables scalable self-supervised pretraining of Mamba-based RS foundation models using large-scale, diverse, unlabeled data. RoMA enhances scalability for high-resolution images through a tailored auto-regressive learning strategy, incorporating two key innovations: 1) a rotation-aware pretraining mechanism combining adaptive cropping with angular embeddings to handle sparsely distributed objects with arbitrary orientations, and 2) multi-scale token prediction objectives that address the extreme variations in object scales inherent to RS imagery. Systematic empirical studies validate that Mamba adheres to RS data and parameter scaling laws, with performance scaling reliably as model and data size increase. Furthermore, experiments across scene classification, object detection, and semantic segmentation tasks demonstrate that RoMA-pretrained Mamba models consistently outperform ViT-based counterparts in both accuracy and computational efficiency. The source code and pretrained models will be released at https://github.com/MiliLab/RoMA.
2503.10398
Niklas Houba
Niklas Houba
Deep source separation of overlapping gravitational-wave signals and non-stationary noise artifacts
19 pages, 10 figures
null
null
null
astro-ph.IM gr-qc physics.data-an physics.ins-det
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The Laser Interferometer Space Antenna (LISA) will observe gravitational waves in the millihertz frequency band, detecting signals from a vast number of astrophysical sources embedded in instrumental noise. Extracting individual signals from these overlapping contributions is a fundamental challenge in LISA data analysis and is traditionally addressed using computationally expensive stochastic Bayesian techniques. In this work, we present a deep learning-based framework for blind source separation in LISA data, employing an encoder-decoder architecture commonly used in digital audio processing to isolate individual signals within complex mixtures. Our approach enables signals from massive black-hole binaries, Galactic binaries, and instrumental glitches to be disentangled directly in a single step, circumventing the need for sequential source identification and subtraction. By learning clustered latent space representations, the framework provides a scalable alternative to conventional methods, with applications in both low-latency event detection and full-scale global-fit analyses. As a proof of concept, we assess the model's performance using simulated LISA data in a controlled setting with a limited number of overlapping sources. The results highlight deep source separation as a promising tool for LISA, paving the way for future extensions to more complex datasets.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 14:19:13 GMT" } ]
2025-03-14T00:00:00
[ [ "Houba", "Niklas", "" ] ]
TITLE: Deep source separation of overlapping gravitational-wave signals and non-stationary noise artifacts ABSTRACT: The Laser Interferometer Space Antenna (LISA) will observe gravitational waves in the millihertz frequency band, detecting signals from a vast number of astrophysical sources embedded in instrumental noise. Extracting individual signals from these overlapping contributions is a fundamental challenge in LISA data analysis and is traditionally addressed using computationally expensive stochastic Bayesian techniques. In this work, we present a deep learning-based framework for blind source separation in LISA data, employing an encoder-decoder architecture commonly used in digital audio processing to isolate individual signals within complex mixtures. Our approach enables signals from massive black-hole binaries, Galactic binaries, and instrumental glitches to be disentangled directly in a single step, circumventing the need for sequential source identification and subtraction. By learning clustered latent space representations, the framework provides a scalable alternative to conventional methods, with applications in both low-latency event detection and full-scale global-fit analyses. As a proof of concept, we assess the model's performance using simulated LISA data in a controlled setting with a limited number of overlapping sources. The results highlight deep source separation as a promising tool for LISA, paving the way for future extensions to more complex datasets.
2503.10404
Fabrizio Pittorino
Matteo Gambella, Fabrizio Pittorino, Manuel Roveri
Architecture-Aware Minimization (A$^2$M): How to Find Flat Minima in Neural Architecture Search
22 pages, 11 figures, 3 tables
null
null
null
cs.LG cond-mat.dis-nn cs.CV
http://creativecommons.org/licenses/by/4.0/
Neural Architecture Search (NAS) has become an essential tool for designing effective and efficient neural networks. In this paper, we investigate the geometric properties of neural architecture spaces commonly used in differentiable NAS methods, specifically NAS-Bench-201 and DARTS. By defining flatness metrics such as neighborhoods and loss barriers along paths in architecture space, we reveal locality and flatness characteristics analogous to the well-known properties of neural network loss landscapes in weight space. In particular, we find that highly accurate architectures cluster together in flat regions, while suboptimal architectures remain isolated, unveiling the detailed geometrical structure of the architecture search landscape. Building on these insights, we propose Architecture-Aware Minimization (A$^2$M), a novel analytically derived algorithmic framework that explicitly biases, for the first time, the gradient of differentiable NAS methods towards flat minima in architecture space. A$^2$M consistently improves generalization over state-of-the-art DARTS-based algorithms on benchmark datasets including CIFAR-10, CIFAR-100, and ImageNet16-120, across both NAS-Bench-201 and DARTS search spaces. Notably, A$^2$M is able to increase the test accuracy, on average across different differentiable NAS methods, by +3.60\% on CIFAR-10, +4.60\% on CIFAR-100, and +3.64\% on ImageNet16-120, demonstrating its superior effectiveness in practice. A$^2$M can be easily integrated into existing differentiable NAS frameworks, offering a versatile tool for future research and applications in automated machine learning. We open-source our code at https://github.com/AI-Tech-Research-Lab/AsquaredM.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 14:30:17 GMT" } ]
2025-03-14T00:00:00
[ [ "Gambella", "Matteo", "" ], [ "Pittorino", "Fabrizio", "" ], [ "Roveri", "Manuel", "" ] ]
TITLE: Architecture-Aware Minimization (A$^2$M): How to Find Flat Minima in Neural Architecture Search ABSTRACT: Neural Architecture Search (NAS) has become an essential tool for designing effective and efficient neural networks. In this paper, we investigate the geometric properties of neural architecture spaces commonly used in differentiable NAS methods, specifically NAS-Bench-201 and DARTS. By defining flatness metrics such as neighborhoods and loss barriers along paths in architecture space, we reveal locality and flatness characteristics analogous to the well-known properties of neural network loss landscapes in weight space. In particular, we find that highly accurate architectures cluster together in flat regions, while suboptimal architectures remain isolated, unveiling the detailed geometrical structure of the architecture search landscape. Building on these insights, we propose Architecture-Aware Minimization (A$^2$M), a novel analytically derived algorithmic framework that explicitly biases, for the first time, the gradient of differentiable NAS methods towards flat minima in architecture space. A$^2$M consistently improves generalization over state-of-the-art DARTS-based algorithms on benchmark datasets including CIFAR-10, CIFAR-100, and ImageNet16-120, across both NAS-Bench-201 and DARTS search spaces. Notably, A$^2$M is able to increase the test accuracy, on average across different differentiable NAS methods, by +3.60\% on CIFAR-10, +4.60\% on CIFAR-100, and +3.64\% on ImageNet16-120, demonstrating its superior effectiveness in practice. A$^2$M can be easily integrated into existing differentiable NAS frameworks, offering a versatile tool for future research and applications in automated machine learning. We open-source our code at https://github.com/AI-Tech-Research-Lab/AsquaredM.
2503.10406
Yijing Lin
Yijing Lin, Mengqi Huang, Shuhan Zhuang, Zhendong Mao
RealGeneral: Unifying Visual Generation via Temporal In-Context Learning with Video Models
null
null
null
null
cs.CV cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Unifying diverse image generation tasks within a single framework remains a fundamental challenge in visual generation. While large language models (LLMs) achieve unification through task-agnostic data and generation, existing visual generation models fail to meet these principles. Current approaches either rely on per-task datasets and large-scale training or adapt pre-trained image models with task-specific modifications, limiting their generalizability. In this work, we explore video models as a foundation for unified image generation, leveraging their inherent ability to model temporal correlations. We introduce RealGeneral, a novel framework that reformulates image generation as a conditional frame prediction task, analogous to in-context learning in LLMs. To bridge the gap between video models and condition-image pairs, we propose (1) a Unified Conditional Embedding module for multi-modal alignment and (2) a Unified Stream DiT Block with decoupled adaptive LayerNorm and attention mask to mitigate cross-modal interference. RealGeneral demonstrates effectiveness in multiple important visual generation tasks, e.g., it achieves a 14.5% improvement in subject similarity for customized generation and a 10% enhancement in image quality for canny-to-image task. Project page: https://lyne1.github.io/RealGeneral/
[ { "version": "v1", "created": "Thu, 13 Mar 2025 14:31:52 GMT" } ]
2025-03-14T00:00:00
[ [ "Lin", "Yijing", "" ], [ "Huang", "Mengqi", "" ], [ "Zhuang", "Shuhan", "" ], [ "Mao", "Zhendong", "" ] ]
TITLE: RealGeneral: Unifying Visual Generation via Temporal In-Context Learning with Video Models ABSTRACT: Unifying diverse image generation tasks within a single framework remains a fundamental challenge in visual generation. While large language models (LLMs) achieve unification through task-agnostic data and generation, existing visual generation models fail to meet these principles. Current approaches either rely on per-task datasets and large-scale training or adapt pre-trained image models with task-specific modifications, limiting their generalizability. In this work, we explore video models as a foundation for unified image generation, leveraging their inherent ability to model temporal correlations. We introduce RealGeneral, a novel framework that reformulates image generation as a conditional frame prediction task, analogous to in-context learning in LLMs. To bridge the gap between video models and condition-image pairs, we propose (1) a Unified Conditional Embedding module for multi-modal alignment and (2) a Unified Stream DiT Block with decoupled adaptive LayerNorm and attention mask to mitigate cross-modal interference. RealGeneral demonstrates effectiveness in multiple important visual generation tasks, e.g., it achieves a 14.5% improvement in subject similarity for customized generation and a 10% enhancement in image quality for canny-to-image task. Project page: https://lyne1.github.io/RealGeneral/
2503.10410
Yuwen Du
Yuwen Du, Anning Hu, Zichen Chao, Yifan Lu, Junhao Ge, Genjia Liu, Weitao Wu, Lanjun Wang, Siheng Chen
RoCo-Sim: Enhancing Roadside Collaborative Perception through Foreground Simulation
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Roadside Collaborative Perception refers to a system where multiple roadside units collaborate to pool their perceptual data, assisting vehicles in enhancing their environmental awareness. Existing roadside perception methods concentrate on model design but overlook data issues like calibration errors, sparse information, and multi-view consistency, leading to poor performance on recent published datasets. To significantly enhance roadside collaborative perception and address critical data issues, we present the first simulation framework RoCo-Sim for road-side collaborative perception. RoCo-Sim is capable of generating diverse, multi-view consistent simulated roadside data through dynamic foreground editing and full-scene style transfer of a single image. RoCo-Sim consists of four components: (1) Camera Extrinsic Optimization ensures accurate 3D to 2D projection for roadside cameras; (2) A novel Multi-View Occlusion-Aware Sampler (MOAS) determines the placement of diverse digital assets within 3D space; (3) DepthSAM innovatively models foreground-background relationships from single-frame fixed-view images, ensuring multi-view consistency of foreground; and (4) Scalable Post-Processing Toolkit generates more realistic and enriched scenes through style transfer and other enhancements. RoCo-Sim significantly improves roadside 3D object detection, outperforming SOTA methods by 83.74 on Rcooper-Intersection and 83.12 on TUMTraf-V2X for AP70. RoCo-Sim fills a critical gap in roadside perception simulation. Code and pre-trained models will be released soon: https://github.com/duyuwen-duen/RoCo-Sim
[ { "version": "v1", "created": "Thu, 13 Mar 2025 14:33:42 GMT" } ]
2025-03-14T00:00:00
[ [ "Du", "Yuwen", "" ], [ "Hu", "Anning", "" ], [ "Chao", "Zichen", "" ], [ "Lu", "Yifan", "" ], [ "Ge", "Junhao", "" ], [ "Liu", "Genjia", "" ], [ "Wu", "Weitao", "" ], [ "Wang", "Lanjun", "" ...
TITLE: RoCo-Sim: Enhancing Roadside Collaborative Perception through Foreground Simulation ABSTRACT: Roadside Collaborative Perception refers to a system where multiple roadside units collaborate to pool their perceptual data, assisting vehicles in enhancing their environmental awareness. Existing roadside perception methods concentrate on model design but overlook data issues like calibration errors, sparse information, and multi-view consistency, leading to poor performance on recent published datasets. To significantly enhance roadside collaborative perception and address critical data issues, we present the first simulation framework RoCo-Sim for road-side collaborative perception. RoCo-Sim is capable of generating diverse, multi-view consistent simulated roadside data through dynamic foreground editing and full-scene style transfer of a single image. RoCo-Sim consists of four components: (1) Camera Extrinsic Optimization ensures accurate 3D to 2D projection for roadside cameras; (2) A novel Multi-View Occlusion-Aware Sampler (MOAS) determines the placement of diverse digital assets within 3D space; (3) DepthSAM innovatively models foreground-background relationships from single-frame fixed-view images, ensuring multi-view consistency of foreground; and (4) Scalable Post-Processing Toolkit generates more realistic and enriched scenes through style transfer and other enhancements. RoCo-Sim significantly improves roadside 3D object detection, outperforming SOTA methods by 83.74 on Rcooper-Intersection and 83.12 on TUMTraf-V2X for AP70. RoCo-Sim fills a critical gap in roadside perception simulation. Code and pre-trained models will be released soon: https://github.com/duyuwen-duen/RoCo-Sim
2503.10421
Zhenwei Wang
Zhenwei Wang, Ruibin Bai, Tiehua Zhang
Towards Constraint-Based Adaptive Hypergraph Learning for Solving Vehicle Routing: An End-to-End Solution
null
null
null
null
cs.LG cs.NE
http://creativecommons.org/licenses/by/4.0/
The application of learning based methods to vehicle routing problems has emerged as a pivotal area of research in combinatorial optimization. These problems are characterized by vast solution spaces and intricate constraints, making traditional approaches such as exact mathematical models or heuristic methods prone to high computational overhead or reliant on the design of complex heuristic operators to achieve optimal or near optimal solutions. Meanwhile, although some recent learning-based methods can produce good performance for VRP with straightforward constraint scenarios, they often fail to effectively handle hard constraints that are common in practice. This study introduces a novel end-to-end framework that combines constraint-oriented hypergraphs with reinforcement learning to address vehicle routing problems. A central innovation of this work is the development of a constraint-oriented dynamic hyperedge reconstruction strategy within an encoder, which significantly enhances hypergraph representation learning. Additionally, the decoder leverages a double-pointer attention mechanism to iteratively generate solutions. The proposed model is trained by incorporating asynchronous parameter updates informed by hypergraph constraints and optimizing a dual loss function comprising constraint loss and policy gradient loss. The experiment results on benchmark datasets demonstrate that the proposed approach not only eliminates the need for sophisticated heuristic operators but also achieves substantial improvements in solution quality.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 14:42:44 GMT" } ]
2025-03-14T00:00:00
[ [ "Wang", "Zhenwei", "" ], [ "Bai", "Ruibin", "" ], [ "Zhang", "Tiehua", "" ] ]
TITLE: Towards Constraint-Based Adaptive Hypergraph Learning for Solving Vehicle Routing: An End-to-End Solution ABSTRACT: The application of learning based methods to vehicle routing problems has emerged as a pivotal area of research in combinatorial optimization. These problems are characterized by vast solution spaces and intricate constraints, making traditional approaches such as exact mathematical models or heuristic methods prone to high computational overhead or reliant on the design of complex heuristic operators to achieve optimal or near optimal solutions. Meanwhile, although some recent learning-based methods can produce good performance for VRP with straightforward constraint scenarios, they often fail to effectively handle hard constraints that are common in practice. This study introduces a novel end-to-end framework that combines constraint-oriented hypergraphs with reinforcement learning to address vehicle routing problems. A central innovation of this work is the development of a constraint-oriented dynamic hyperedge reconstruction strategy within an encoder, which significantly enhances hypergraph representation learning. Additionally, the decoder leverages a double-pointer attention mechanism to iteratively generate solutions. The proposed model is trained by incorporating asynchronous parameter updates informed by hypergraph constraints and optimizing a dual loss function comprising constraint loss and policy gradient loss. The experiment results on benchmark datasets demonstrate that the proposed approach not only eliminates the need for sophisticated heuristic operators but also achieves substantial improvements in solution quality.
2503.10426
Javad Pourmostafa Roshan Sharami
Bennet van den Broek and Javad Pourmostafa Roshan Sharami
Improving Medical Waste Classification with Hybrid Capsule Networks
null
null
null
null
cs.CV cs.LG
http://creativecommons.org/licenses/by/4.0/
The improper disposal and mismanagement of medical waste pose severe environmental and public health risks, contributing to greenhouse gas emissions and the spread of infectious diseases. Efficient and accurate medical waste classification is crucial for mitigating these risks. We explore the integration of capsule networks with a pretrained DenseNet model to improve medical waste classification. To the best of our knowledge, capsule networks have not yet been applied to this task, making this study the first to assess their effectiveness. A diverse dataset of medical waste images collected from multiple public sources, is used to evaluate three model configurations: (1) a pretrained DenseNet model as a baseline, (2) a pretrained DenseNet with frozen layers combined with a capsule network, and (3) a pretrained DenseNet with unfrozen layers combined with a capsule network. Experimental results demonstrate that incorporating capsule networks improves classification performance, with F1 scores increasing from 0.89 (baseline) to 0.92 (hybrid model with unfrozen layers). This highlights the potential of capsule networks to address the spatial limitations of traditional convolutional models and improve classification robustness. While the capsule-enhanced model demonstrated improved classification performance, direct comparisons with prior studies were challenging due to differences in dataset size and diversity. Previous studies relied on smaller, domain-specific datasets, which inherently yielded higher accuracy. In contrast, our study employs a significantly larger and more diverse dataset, leading to better generalization but introducing additional classification challenges. This highlights the trade-off between dataset complexity and model performance.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 14:49:30 GMT" } ]
2025-03-14T00:00:00
[ [ "Broek", "Bennet van den", "" ], [ "Sharami", "Javad Pourmostafa Roshan", "" ] ]
TITLE: Improving Medical Waste Classification with Hybrid Capsule Networks ABSTRACT: The improper disposal and mismanagement of medical waste pose severe environmental and public health risks, contributing to greenhouse gas emissions and the spread of infectious diseases. Efficient and accurate medical waste classification is crucial for mitigating these risks. We explore the integration of capsule networks with a pretrained DenseNet model to improve medical waste classification. To the best of our knowledge, capsule networks have not yet been applied to this task, making this study the first to assess their effectiveness. A diverse dataset of medical waste images collected from multiple public sources, is used to evaluate three model configurations: (1) a pretrained DenseNet model as a baseline, (2) a pretrained DenseNet with frozen layers combined with a capsule network, and (3) a pretrained DenseNet with unfrozen layers combined with a capsule network. Experimental results demonstrate that incorporating capsule networks improves classification performance, with F1 scores increasing from 0.89 (baseline) to 0.92 (hybrid model with unfrozen layers). This highlights the potential of capsule networks to address the spatial limitations of traditional convolutional models and improve classification robustness. While the capsule-enhanced model demonstrated improved classification performance, direct comparisons with prior studies were challenging due to differences in dataset size and diversity. Previous studies relied on smaller, domain-specific datasets, which inherently yielded higher accuracy. In contrast, our study employs a significantly larger and more diverse dataset, leading to better generalization but introducing additional classification challenges. This highlights the trade-off between dataset complexity and model performance.
2503.10434
Derun Li
Derun Li, Jianwei Ren, Yue Wang, Xin Wen, Pengxiang Li, Leimeng Xu, Kun Zhan, Zhongpu Xia, Peng Jia, Xianpeng Lang, Ningyi Xu, Hang Zhao
Finetuning Generative Trajectory Model with Reinforcement Learning from Human Feedback
10 pages, 5 figures
null
null
null
cs.RO cs.CV cs.LG
http://creativecommons.org/licenses/by/4.0/
Generating human-like and adaptive trajectories is essential for autonomous driving in dynamic environments. While generative models have shown promise in synthesizing feasible trajectories, they often fail to capture the nuanced variability of human driving styles due to dataset biases and distributional shifts. To address this, we introduce TrajHF, a human feedback-driven finetuning framework for generative trajectory models, designed to align motion planning with diverse driving preferences. TrajHF incorporates multi-conditional denoiser and reinforcement learning with human feedback to refine multi-modal trajectory generation beyond conventional imitation learning. This enables better alignment with human driving preferences while maintaining safety and feasibility constraints. TrajHF achieves PDMS of 93.95 on NavSim benchmark, significantly exceeding other methods. TrajHF sets a new paradigm for personalized and adaptable trajectory generation in autonomous driving.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 14:56:17 GMT" } ]
2025-03-14T00:00:00
[ [ "Li", "Derun", "" ], [ "Ren", "Jianwei", "" ], [ "Wang", "Yue", "" ], [ "Wen", "Xin", "" ], [ "Li", "Pengxiang", "" ], [ "Xu", "Leimeng", "" ], [ "Zhan", "Kun", "" ], [ "Xia", "Zhongpu", "" ...
TITLE: Finetuning Generative Trajectory Model with Reinforcement Learning from Human Feedback ABSTRACT: Generating human-like and adaptive trajectories is essential for autonomous driving in dynamic environments. While generative models have shown promise in synthesizing feasible trajectories, they often fail to capture the nuanced variability of human driving styles due to dataset biases and distributional shifts. To address this, we introduce TrajHF, a human feedback-driven finetuning framework for generative trajectory models, designed to align motion planning with diverse driving preferences. TrajHF incorporates multi-conditional denoiser and reinforcement learning with human feedback to refine multi-modal trajectory generation beyond conventional imitation learning. This enables better alignment with human driving preferences while maintaining safety and feasibility constraints. TrajHF achieves PDMS of 93.95 on NavSim benchmark, significantly exceeding other methods. TrajHF sets a new paradigm for personalized and adaptable trajectory generation in autonomous driving.
2503.10450
Maarten Perneel
Maarten Perneel, Ines Adriaens, Ben Aernouts, Jan Verwaeren
Consistent multi-animal pose estimation in cattle using dynamic Kalman filter based tracking
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by-sa/4.0/
Over the past decade, studying animal behaviour with the help of computer vision has become more popular. Replacing human observers by computer vision lowers the cost of data collection and therefore allows to collect more extensive datasets. However, the majority of available computer vision algorithms to study animal behaviour is highly tailored towards a single research objective, limiting possibilities for data reuse. In this perspective, pose-estimation in combination with animal tracking offers opportunities to yield a higher level representation capturing both the spatial and temporal component of animal behaviour. Such a higher level representation allows to answer a wide variety of research questions simultaneously, without the need to develop repeatedly tailored computer vision algorithms. In this paper, we therefore first cope with several weaknesses of current pose-estimation algorithms and thereafter introduce KeySORT (Keypoint Simple and Online Realtime Tracking). KeySORT deploys an adaptive Kalman filter to construct tracklets in a bounding-box free manner, significantly improving the temporal consistency of detected keypoints. In this paper, we focus on pose estimation in cattle, but our methodology can easily be generalised to any other animal species. Our test results indicate our algorithm is able to detect up to 80% of the ground truth keypoints with high accuracy, with only a limited drop in performance when daylight recordings are compared to nightvision recordings. Moreover, by using KeySORT to construct skeletons, the temporal consistency of generated keypoint coordinates was largely improved, offering opportunities with regard to automated behaviour monitoring of animals.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 15:15:54 GMT" } ]
2025-03-14T00:00:00
[ [ "Perneel", "Maarten", "" ], [ "Adriaens", "Ines", "" ], [ "Aernouts", "Ben", "" ], [ "Verwaeren", "Jan", "" ] ]
TITLE: Consistent multi-animal pose estimation in cattle using dynamic Kalman filter based tracking ABSTRACT: Over the past decade, studying animal behaviour with the help of computer vision has become more popular. Replacing human observers by computer vision lowers the cost of data collection and therefore allows to collect more extensive datasets. However, the majority of available computer vision algorithms to study animal behaviour is highly tailored towards a single research objective, limiting possibilities for data reuse. In this perspective, pose-estimation in combination with animal tracking offers opportunities to yield a higher level representation capturing both the spatial and temporal component of animal behaviour. Such a higher level representation allows to answer a wide variety of research questions simultaneously, without the need to develop repeatedly tailored computer vision algorithms. In this paper, we therefore first cope with several weaknesses of current pose-estimation algorithms and thereafter introduce KeySORT (Keypoint Simple and Online Realtime Tracking). KeySORT deploys an adaptive Kalman filter to construct tracklets in a bounding-box free manner, significantly improving the temporal consistency of detected keypoints. In this paper, we focus on pose estimation in cattle, but our methodology can easily be generalised to any other animal species. Our test results indicate our algorithm is able to detect up to 80% of the ground truth keypoints with high accuracy, with only a limited drop in performance when daylight recordings are compared to nightvision recordings. Moreover, by using KeySORT to construct skeletons, the temporal consistency of generated keypoint coordinates was largely improved, offering opportunities with regard to automated behaviour monitoring of animals.
2503.10452
Wenhao Hu
Wenhao Hu, Jinhao Duan, Chunchen Wei, Li Zhang, Yue Zhang, Kaidi Xu
DynaCode: A Dynamic Complexity-Aware Code Benchmark for Evaluating Large Language Models in Code Generation
16 pages, 11 figures
null
null
null
cs.CL cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The rapid advancement of large language models (LLMs) has significantly improved their performance in code generation tasks. However, existing code benchmarks remain static, consisting of fixed datasets with predefined problems. This makes them vulnerable to memorization during training, where LLMs recall specific test cases instead of generalizing to new problems, leading to data contamination and unreliable evaluation results. To address these issues, we introduce DynaCode, a dynamic, complexity-aware benchmark that overcomes the limitations of static datasets. DynaCode evaluates LLMs systematically using a complexity-aware metric, incorporating both code complexity and call-graph structures. DynaCode achieves large-scale diversity, generating up to 189 million unique nested code problems across four distinct levels of code complexity, referred to as units, and 16 types of call graphs. Results on 12 latest LLMs show an average performance drop of 16.8% to 45.7% compared to MBPP+, a static code generation benchmark, with performance progressively decreasing as complexity increases. This demonstrates DynaCode's ability to effectively differentiate LLMs. Additionally, by leveraging call graphs, we gain insights into LLM behavior, particularly their preference for handling subfunction interactions within nested code.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 15:18:56 GMT" } ]
2025-03-14T00:00:00
[ [ "Hu", "Wenhao", "" ], [ "Duan", "Jinhao", "" ], [ "Wei", "Chunchen", "" ], [ "Zhang", "Li", "" ], [ "Zhang", "Yue", "" ], [ "Xu", "Kaidi", "" ] ]
TITLE: DynaCode: A Dynamic Complexity-Aware Code Benchmark for Evaluating Large Language Models in Code Generation ABSTRACT: The rapid advancement of large language models (LLMs) has significantly improved their performance in code generation tasks. However, existing code benchmarks remain static, consisting of fixed datasets with predefined problems. This makes them vulnerable to memorization during training, where LLMs recall specific test cases instead of generalizing to new problems, leading to data contamination and unreliable evaluation results. To address these issues, we introduce DynaCode, a dynamic, complexity-aware benchmark that overcomes the limitations of static datasets. DynaCode evaluates LLMs systematically using a complexity-aware metric, incorporating both code complexity and call-graph structures. DynaCode achieves large-scale diversity, generating up to 189 million unique nested code problems across four distinct levels of code complexity, referred to as units, and 16 types of call graphs. Results on 12 latest LLMs show an average performance drop of 16.8% to 45.7% compared to MBPP+, a static code generation benchmark, with performance progressively decreasing as complexity increases. This demonstrates DynaCode's ability to effectively differentiate LLMs. Additionally, by leveraging call graphs, we gain insights into LLM behavior, particularly their preference for handling subfunction interactions within nested code.
2503.10464
Xunzhi Zheng
Xunzhi Zheng and Dan Xu
Flow-NeRF: Joint Learning of Geometry, Poses, and Dense Flow within Unified Neural Representations
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Learning accurate scene reconstruction without pose priors in neural radiance fields is challenging due to inherent geometric ambiguity. Recent development either relies on correspondence priors for regularization or uses off-the-shelf flow estimators to derive analytical poses. However, the potential for jointly learning scene geometry, camera poses, and dense flow within a unified neural representation remains largely unexplored. In this paper, we present Flow-NeRF, a unified framework that simultaneously optimizes scene geometry, camera poses, and dense optical flow all on-the-fly. To enable the learning of dense flow within the neural radiance field, we design and build a bijective mapping for flow estimation, conditioned on pose. To make the scene reconstruction benefit from the flow estimation, we develop an effective feature enhancement mechanism to pass canonical space features to world space representations, significantly enhancing scene geometry. We validate our model across four important tasks, i.e., novel view synthesis, depth estimation, camera pose prediction, and dense optical flow estimation, using several datasets. Our approach surpasses previous methods in almost all metrics for novel-view view synthesis and depth estimation and yields both qualitatively sound and quantitatively accurate novel-view flow. Our project page is https://zhengxunzhi.github.io/flownerf/.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 15:37:11 GMT" } ]
2025-03-14T00:00:00
[ [ "Zheng", "Xunzhi", "" ], [ "Xu", "Dan", "" ] ]
TITLE: Flow-NeRF: Joint Learning of Geometry, Poses, and Dense Flow within Unified Neural Representations ABSTRACT: Learning accurate scene reconstruction without pose priors in neural radiance fields is challenging due to inherent geometric ambiguity. Recent development either relies on correspondence priors for regularization or uses off-the-shelf flow estimators to derive analytical poses. However, the potential for jointly learning scene geometry, camera poses, and dense flow within a unified neural representation remains largely unexplored. In this paper, we present Flow-NeRF, a unified framework that simultaneously optimizes scene geometry, camera poses, and dense optical flow all on-the-fly. To enable the learning of dense flow within the neural radiance field, we design and build a bijective mapping for flow estimation, conditioned on pose. To make the scene reconstruction benefit from the flow estimation, we develop an effective feature enhancement mechanism to pass canonical space features to world space representations, significantly enhancing scene geometry. We validate our model across four important tasks, i.e., novel view synthesis, depth estimation, camera pose prediction, and dense optical flow estimation, using several datasets. Our approach surpasses previous methods in almost all metrics for novel-view view synthesis and depth estimation and yields both qualitatively sound and quantitatively accurate novel-view flow. Our project page is https://zhengxunzhi.github.io/flownerf/.
2503.10471
Liming Wu
Liming Wu, Wenbing Huang, Rui Jiao, Jianxing Huang, Liwei Liu, Yipeng Zhou, Hao Sun, Yang Liu, Fuchun Sun, Yuxiang Ren, Jirong Wen
Siamese Foundation Models for Crystal Structure Prediction
null
null
null
null
cond-mat.mtrl-sci cs.AI
http://creativecommons.org/licenses/by/4.0/
Crystal Structure Prediction (CSP), which aims to generate stable crystal structures from compositions, represents a critical pathway for discovering novel materials. While structure prediction tasks in other domains, such as proteins, have seen remarkable progress, CSP remains a relatively underexplored area due to the more complex geometries inherent in crystal structures. In this paper, we propose Siamese foundation models specifically designed to address CSP. Our pretrain-finetune framework, named DAO, comprises two complementary foundation models: DAO-G for structure generation and DAO-P for energy prediction. Experiments on CSP benchmarks (MP-20 and MPTS-52) demonstrate that our DAO-G significantly surpasses state-of-the-art (SOTA) methods across all metrics. Extensive ablation studies further confirm that DAO-G excels in generating diverse polymorphic structures, and the dataset relaxation and energy guidance provided by DAO-P are essential for enhancing DAO-G's performance. When applied to three real-world superconductors ($\text{CsV}_3\text{Sb}_5$, $ \text{Zr}_{16}\text{Rh}_8\text{O}_4$ and $\text{Zr}_{16}\text{Pd}_8\text{O}_4$) that are known to be challenging to analyze, our foundation models achieve accurate critical temperature predictions and structure generations. For instance, on $\text{CsV}_3\text{Sb}_5$, DAO-G generates a structure close to the experimental one with an RMSE of 0.0085; DAO-P predicts the $T_c$ value with high accuracy (2.26 K vs. the ground-truth value of 2.30 K). In contrast, conventional DFT calculators like Quantum Espresso only successfully derive the structure of the first superconductor within an acceptable time, while the RMSE is nearly 8 times larger, and the computation speed is more than 1000 times slower. These compelling results collectively highlight the potential of our approach for advancing materials science research and development.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 15:44:16 GMT" } ]
2025-03-14T00:00:00
[ [ "Wu", "Liming", "" ], [ "Huang", "Wenbing", "" ], [ "Jiao", "Rui", "" ], [ "Huang", "Jianxing", "" ], [ "Liu", "Liwei", "" ], [ "Zhou", "Yipeng", "" ], [ "Sun", "Hao", "" ], [ "Liu", "Yang", ...
TITLE: Siamese Foundation Models for Crystal Structure Prediction ABSTRACT: Crystal Structure Prediction (CSP), which aims to generate stable crystal structures from compositions, represents a critical pathway for discovering novel materials. While structure prediction tasks in other domains, such as proteins, have seen remarkable progress, CSP remains a relatively underexplored area due to the more complex geometries inherent in crystal structures. In this paper, we propose Siamese foundation models specifically designed to address CSP. Our pretrain-finetune framework, named DAO, comprises two complementary foundation models: DAO-G for structure generation and DAO-P for energy prediction. Experiments on CSP benchmarks (MP-20 and MPTS-52) demonstrate that our DAO-G significantly surpasses state-of-the-art (SOTA) methods across all metrics. Extensive ablation studies further confirm that DAO-G excels in generating diverse polymorphic structures, and the dataset relaxation and energy guidance provided by DAO-P are essential for enhancing DAO-G's performance. When applied to three real-world superconductors ($\text{CsV}_3\text{Sb}_5$, $ \text{Zr}_{16}\text{Rh}_8\text{O}_4$ and $\text{Zr}_{16}\text{Pd}_8\text{O}_4$) that are known to be challenging to analyze, our foundation models achieve accurate critical temperature predictions and structure generations. For instance, on $\text{CsV}_3\text{Sb}_5$, DAO-G generates a structure close to the experimental one with an RMSE of 0.0085; DAO-P predicts the $T_c$ value with high accuracy (2.26 K vs. the ground-truth value of 2.30 K). In contrast, conventional DFT calculators like Quantum Espresso only successfully derive the structure of the first superconductor within an acceptable time, while the RMSE is nearly 8 times larger, and the computation speed is more than 1000 times slower. These compelling results collectively highlight the potential of our approach for advancing materials science research and development.
2503.10486
Gaurav Kumar Gupta
Gaurav Kumar Gupta and Pranal Pande
LLMs in Disease Diagnosis: A Comparative Study of DeepSeek-R1 and O3 Mini Across Chronic Health Conditions
12 pages, 3 figures
null
null
null
cs.CL cs.AI
http://creativecommons.org/licenses/by/4.0/
Large Language Models (LLMs) are revolutionizing medical diagnostics by enhancing both disease classification and clinical decision-making. In this study, we evaluate the performance of two LLM- based diagnostic tools, DeepSeek R1 and O3 Mini, using a structured dataset of symptoms and diagnoses. We assessed their predictive accuracy at both the disease and category levels, as well as the reliability of their confidence scores. DeepSeek R1 achieved a disease-level accuracy of 76% and an overall accuracy of 82%, outperforming O3 Mini, which attained 72% and 75% respectively. Notably, DeepSeek R1 demonstrated exceptional performance in Mental Health, Neurological Disorders, and Oncology, where it reached 100% accuracy, while O3 Mini excelled in Autoimmune Disease classification with 100% accuracy. Both models, however, struggled with Respiratory Disease classification, recording accuracies of only 40% for DeepSeek R1 and 20% for O3 Mini. Additionally, the analysis of confidence scores revealed that DeepSeek R1 provided high-confidence predictions in 92% of cases, compared to 68% for O3 Mini. Ethical considerations regarding bias, model interpretability, and data privacy are also discussed to ensure the responsible integration of LLMs into clinical practice. Overall, our findings offer valuable insights into the strengths and limitations of LLM-based diagnostic systems and provide a roadmap for future enhancements in AI-driven healthcare.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 15:54:26 GMT" } ]
2025-03-14T00:00:00
[ [ "Gupta", "Gaurav Kumar", "" ], [ "Pande", "Pranal", "" ] ]
TITLE: LLMs in Disease Diagnosis: A Comparative Study of DeepSeek-R1 and O3 Mini Across Chronic Health Conditions ABSTRACT: Large Language Models (LLMs) are revolutionizing medical diagnostics by enhancing both disease classification and clinical decision-making. In this study, we evaluate the performance of two LLM- based diagnostic tools, DeepSeek R1 and O3 Mini, using a structured dataset of symptoms and diagnoses. We assessed their predictive accuracy at both the disease and category levels, as well as the reliability of their confidence scores. DeepSeek R1 achieved a disease-level accuracy of 76% and an overall accuracy of 82%, outperforming O3 Mini, which attained 72% and 75% respectively. Notably, DeepSeek R1 demonstrated exceptional performance in Mental Health, Neurological Disorders, and Oncology, where it reached 100% accuracy, while O3 Mini excelled in Autoimmune Disease classification with 100% accuracy. Both models, however, struggled with Respiratory Disease classification, recording accuracies of only 40% for DeepSeek R1 and 20% for O3 Mini. Additionally, the analysis of confidence scores revealed that DeepSeek R1 provided high-confidence predictions in 92% of cases, compared to 68% for O3 Mini. Ethical considerations regarding bias, model interpretability, and data privacy are also discussed to ensure the responsible integration of LLMs into clinical practice. Overall, our findings offer valuable insights into the strengths and limitations of LLM-based diagnostic systems and provide a roadmap for future enhancements in AI-driven healthcare.
2503.10501
Xudong Tan
Xudong Tan, Peng Ye, Chongjun Tu, Jianjian Cao, Yaoxin Yang, Lin Zhang, Dongzhan Zhou, Tao Chen
TokenCarve: Information-Preserving Visual Token Compression in Multimodal Large Language Models
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
Multimodal Large Language Models (MLLMs) are becoming increasingly popular, while the high computational cost associated with multimodal data input, particularly from visual tokens, poses a significant challenge. Existing training-based token compression methods improve inference efficiency but require costly retraining, while training-free methods struggle to maintain performance when aggressively reducing token counts. In this study, we reveal that the performance degradation of MLLM closely correlates with the accelerated loss of information in the attention output matrix. This insight introduces a novel information-preserving perspective, making it possible to maintain performance even under extreme token compression. Based on this finding, we propose TokenCarve, a training-free, plug-and-play, two-stage token compression framework. The first stage employs an Information-Preservation-Guided Selection (IPGS) strategy to prune low-information tokens, while the second stage further leverages IPGS to guide token merging, minimizing information loss. Extensive experiments on 11 datasets and 2 model variants demonstrate the effectiveness of TokenCarve. It can even reduce the number of visual tokens to 22.2% of the original count, achieving a 1.23x speedup in inference, a 64% reduction in KV cache storage, and only a 1.54% drop in accuracy. Our code is available at https://github.com/ShawnTan86/TokenCarve.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 16:04:31 GMT" } ]
2025-03-14T00:00:00
[ [ "Tan", "Xudong", "" ], [ "Ye", "Peng", "" ], [ "Tu", "Chongjun", "" ], [ "Cao", "Jianjian", "" ], [ "Yang", "Yaoxin", "" ], [ "Zhang", "Lin", "" ], [ "Zhou", "Dongzhan", "" ], [ "Chen", "Tao", ...
TITLE: TokenCarve: Information-Preserving Visual Token Compression in Multimodal Large Language Models ABSTRACT: Multimodal Large Language Models (MLLMs) are becoming increasingly popular, while the high computational cost associated with multimodal data input, particularly from visual tokens, poses a significant challenge. Existing training-based token compression methods improve inference efficiency but require costly retraining, while training-free methods struggle to maintain performance when aggressively reducing token counts. In this study, we reveal that the performance degradation of MLLM closely correlates with the accelerated loss of information in the attention output matrix. This insight introduces a novel information-preserving perspective, making it possible to maintain performance even under extreme token compression. Based on this finding, we propose TokenCarve, a training-free, plug-and-play, two-stage token compression framework. The first stage employs an Information-Preservation-Guided Selection (IPGS) strategy to prune low-information tokens, while the second stage further leverages IPGS to guide token merging, minimizing information loss. Extensive experiments on 11 datasets and 2 model variants demonstrate the effectiveness of TokenCarve. It can even reduce the number of visual tokens to 22.2% of the original count, achieving a 1.23x speedup in inference, a 64% reduction in KV cache storage, and only a 1.54% drop in accuracy. Our code is available at https://github.com/ShawnTan86/TokenCarve.
2503.10512
Devjeet Roy
Hooman Shahrokhi, Devjeet Raj Roy, Yan Yan, Venera Arnaoudova and Janaradhan Rao Doppa
Conformal Prediction Sets for Deep Generative Models via Reduction to Conformal Regression
null
null
null
null
cs.LG cs.AI
http://creativecommons.org/licenses/by-sa/4.0/
We consider the problem of generating valid and small prediction sets by sampling outputs (e.g., software code and natural language text) from a black-box deep generative model for a given input (e.g., textual prompt). The validity of a prediction set is determined by a user-defined binary admissibility function depending on the target application. For example, requiring at least one program in the set to pass all test cases in code generation application. To address this problem, we develop a simple and effective conformal inference algorithm referred to as Generative Prediction Sets (GPS). Given a set of calibration examples and black-box access to a deep generative model, GPS can generate prediction sets with provable guarantees. The key insight behind GPS is to exploit the inherent structure within the distribution over the minimum number of samples needed to obtain an admissible output to develop a simple conformal regression approach over the minimum number of samples. Experiments on multiple datasets for code and math word problems using different large language models demonstrate the efficacy of GPS over state-of-the-art methods.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 16:16:23 GMT" } ]
2025-03-14T00:00:00
[ [ "Shahrokhi", "Hooman", "" ], [ "Roy", "Devjeet Raj", "" ], [ "Yan", "Yan", "" ], [ "Arnaoudova", "Venera", "" ], [ "Doppa", "Janaradhan Rao", "" ] ]
TITLE: Conformal Prediction Sets for Deep Generative Models via Reduction to Conformal Regression ABSTRACT: We consider the problem of generating valid and small prediction sets by sampling outputs (e.g., software code and natural language text) from a black-box deep generative model for a given input (e.g., textual prompt). The validity of a prediction set is determined by a user-defined binary admissibility function depending on the target application. For example, requiring at least one program in the set to pass all test cases in code generation application. To address this problem, we develop a simple and effective conformal inference algorithm referred to as Generative Prediction Sets (GPS). Given a set of calibration examples and black-box access to a deep generative model, GPS can generate prediction sets with provable guarantees. The key insight behind GPS is to exploit the inherent structure within the distribution over the minimum number of samples needed to obtain an admissible output to develop a simple conformal regression approach over the minimum number of samples. Experiments on multiple datasets for code and math word problems using different large language models demonstrate the efficacy of GPS over state-of-the-art methods.
2503.10522
Zeyue Tian
Zeyue Tian, Yizhu Jin, Zhaoyang Liu, Ruibin Yuan, Xu Tan, Qifeng Chen, Wei Xue, Yike Guo
AudioX: Diffusion Transformer for Anything-to-Audio Generation
The code and datasets will be available at https://zeyuet.github.io/AudioX/
null
null
null
cs.MM cs.CV cs.LG cs.SD eess.AS
http://creativecommons.org/licenses/by/4.0/
Audio and music generation have emerged as crucial tasks in many applications, yet existing approaches face significant limitations: they operate in isolation without unified capabilities across modalities, suffer from scarce high-quality, multi-modal training data, and struggle to effectively integrate diverse inputs. In this work, we propose AudioX, a unified Diffusion Transformer model for Anything-to-Audio and Music Generation. Unlike previous domain-specific models, AudioX can generate both general audio and music with high quality, while offering flexible natural language control and seamless processing of various modalities including text, video, image, music, and audio. Its key innovation is a multi-modal masked training strategy that masks inputs across modalities and forces the model to learn from masked inputs, yielding robust and unified cross-modal representations. To address data scarcity, we curate two comprehensive datasets: vggsound-caps with 190K audio captions based on the VGGSound dataset, and V2M-caps with 6 million music captions derived from the V2M dataset. Extensive experiments demonstrate that AudioX not only matches or outperforms state-of-the-art specialized models, but also offers remarkable versatility in handling diverse input modalities and generation tasks within a unified architecture. The code and datasets will be available at https://zeyuet.github.io/AudioX/
[ { "version": "v1", "created": "Thu, 13 Mar 2025 16:30:59 GMT" } ]
2025-03-14T00:00:00
[ [ "Tian", "Zeyue", "" ], [ "Jin", "Yizhu", "" ], [ "Liu", "Zhaoyang", "" ], [ "Yuan", "Ruibin", "" ], [ "Tan", "Xu", "" ], [ "Chen", "Qifeng", "" ], [ "Xue", "Wei", "" ], [ "Guo", "Yike", "" ...
TITLE: AudioX: Diffusion Transformer for Anything-to-Audio Generation ABSTRACT: Audio and music generation have emerged as crucial tasks in many applications, yet existing approaches face significant limitations: they operate in isolation without unified capabilities across modalities, suffer from scarce high-quality, multi-modal training data, and struggle to effectively integrate diverse inputs. In this work, we propose AudioX, a unified Diffusion Transformer model for Anything-to-Audio and Music Generation. Unlike previous domain-specific models, AudioX can generate both general audio and music with high quality, while offering flexible natural language control and seamless processing of various modalities including text, video, image, music, and audio. Its key innovation is a multi-modal masked training strategy that masks inputs across modalities and forces the model to learn from masked inputs, yielding robust and unified cross-modal representations. To address data scarcity, we curate two comprehensive datasets: vggsound-caps with 190K audio captions based on the VGGSound dataset, and V2M-caps with 6 million music captions derived from the V2M dataset. Extensive experiments demonstrate that AudioX not only matches or outperforms state-of-the-art specialized models, but also offers remarkable versatility in handling diverse input modalities and generation tasks within a unified architecture. The code and datasets will be available at https://zeyuet.github.io/AudioX/
2503.10523
Yongqi Wang
Jun Yu, Yongqi Wang, Lei Wang, Yang Zheng, Shengfan Xu
Interactive Multimodal Fusion with Temporal Modeling
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
This paper presents our method for the estimation of valence-arousal (VA) in the 8th Affective Behavior Analysis in-the-Wild (ABAW) competition. Our approach integrates visual and audio information through a multimodal framework. The visual branch uses a pre-trained ResNet model to extract spatial features from facial images. The audio branches employ pre-trained VGG models to extract VGGish and LogMel features from speech signals. These features undergo temporal modeling using Temporal Convolutional Networks (TCNs). We then apply cross-modal attention mechanisms, where visual features interact with audio features through query-key-value attention structures. Finally, the features are concatenated and passed through a regression layer to predict valence and arousal. Our method achieves competitive performance on the Aff-Wild2 dataset, demonstrating effective multimodal fusion for VA estimation in-the-wild.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 16:31:56 GMT" } ]
2025-03-14T00:00:00
[ [ "Yu", "Jun", "" ], [ "Wang", "Yongqi", "" ], [ "Wang", "Lei", "" ], [ "Zheng", "Yang", "" ], [ "Xu", "Shengfan", "" ] ]
TITLE: Interactive Multimodal Fusion with Temporal Modeling ABSTRACT: This paper presents our method for the estimation of valence-arousal (VA) in the 8th Affective Behavior Analysis in-the-Wild (ABAW) competition. Our approach integrates visual and audio information through a multimodal framework. The visual branch uses a pre-trained ResNet model to extract spatial features from facial images. The audio branches employ pre-trained VGG models to extract VGGish and LogMel features from speech signals. These features undergo temporal modeling using Temporal Convolutional Networks (TCNs). We then apply cross-modal attention mechanisms, where visual features interact with audio features through query-key-value attention structures. Finally, the features are concatenated and passed through a regression layer to predict valence and arousal. Our method achieves competitive performance on the Aff-Wild2 dataset, demonstrating effective multimodal fusion for VA estimation in-the-wild.
2503.10529
Zilu Guo
Zilu Guo, Hongbin Lin, Zhihao Yuan, Chaoda Zheng, Pengshuo Qiu, Dongzhi Jiang, Renrui Zhang, Chun-Mei Feng, Zhen Li
PiSA: A Self-Augmented Data Engine and Training Strategy for 3D Understanding with Large Models
Technical Report
null
null
null
cs.CV cs.AI
http://creativecommons.org/licenses/by/4.0/
3D Multimodal Large Language Models (MLLMs) have recently made substantial advancements. However, their potential remains untapped, primarily due to the limited quantity and suboptimal quality of 3D datasets. Current approaches attempt to transfer knowledge from 2D MLLMs to expand 3D instruction data, but still face modality and domain gaps. To this end, we introduce PiSA-Engine (Point-Self-Augmented-Engine), a new framework for generating instruction point-language datasets enriched with 3D spatial semantics. We observe that existing 3D MLLMs offer a comprehensive understanding of point clouds for annotation, while 2D MLLMs excel at cross-validation by providing complementary information. By integrating holistic 2D and 3D insights from off-the-shelf MLLMs, PiSA-Engine enables a continuous cycle of high-quality data generation. We select PointLLM as the baseline and adopt this co-evolution training framework to develop an enhanced 3D MLLM, termed PointLLM-PiSA. Additionally, we identify limitations in previous 3D benchmarks, which often feature coarse language captions and insufficient category diversity, resulting in inaccurate evaluations. To address this gap, we further introduce PiSA-Bench, a comprehensive 3D benchmark covering six key aspects with detailed and diverse labels. Experimental results demonstrate PointLLM-PiSA's state-of-the-art performance in zero-shot 3D object captioning and generative classification on our PiSA-Bench, achieving significant improvements of 46.45% (+8.33%) and 63.75% (+16.25%), respectively. We will release the code, datasets, and benchmark.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 16:37:26 GMT" } ]
2025-03-14T00:00:00
[ [ "Guo", "Zilu", "" ], [ "Lin", "Hongbin", "" ], [ "Yuan", "Zhihao", "" ], [ "Zheng", "Chaoda", "" ], [ "Qiu", "Pengshuo", "" ], [ "Jiang", "Dongzhi", "" ], [ "Zhang", "Renrui", "" ], [ "Feng", "C...
TITLE: PiSA: A Self-Augmented Data Engine and Training Strategy for 3D Understanding with Large Models ABSTRACT: 3D Multimodal Large Language Models (MLLMs) have recently made substantial advancements. However, their potential remains untapped, primarily due to the limited quantity and suboptimal quality of 3D datasets. Current approaches attempt to transfer knowledge from 2D MLLMs to expand 3D instruction data, but still face modality and domain gaps. To this end, we introduce PiSA-Engine (Point-Self-Augmented-Engine), a new framework for generating instruction point-language datasets enriched with 3D spatial semantics. We observe that existing 3D MLLMs offer a comprehensive understanding of point clouds for annotation, while 2D MLLMs excel at cross-validation by providing complementary information. By integrating holistic 2D and 3D insights from off-the-shelf MLLMs, PiSA-Engine enables a continuous cycle of high-quality data generation. We select PointLLM as the baseline and adopt this co-evolution training framework to develop an enhanced 3D MLLM, termed PointLLM-PiSA. Additionally, we identify limitations in previous 3D benchmarks, which often feature coarse language captions and insufficient category diversity, resulting in inaccurate evaluations. To address this gap, we further introduce PiSA-Bench, a comprehensive 3D benchmark covering six key aspects with detailed and diverse labels. Experimental results demonstrate PointLLM-PiSA's state-of-the-art performance in zero-shot 3D object captioning and generative classification on our PiSA-Bench, achieving significant improvements of 46.45% (+8.33%) and 63.75% (+16.25%), respectively. We will release the code, datasets, and benchmark.
2503.10539
Reshma Rastogi
Reshma Rastogi and Ankush Bisht and Sanjay Kumar and Suresh Chandra
GBSVR: Granular Ball Support Vector Regression
null
null
null
null
cs.LG cs.AI cs.IR
http://creativecommons.org/licenses/by/4.0/
Support Vector Regression (SVR) and its variants are widely used to handle regression tasks, however, since their solution involves solving an expensive quadratic programming problem, it limits its application, especially when dealing with large datasets. Additionally, SVR uses an epsilon-insensitive loss function which is sensitive to outliers and therefore can adversely affect its performance. We propose Granular Ball Support Vector Regression (GBSVR) to tackle problem of regression by using granular ball concept. These balls are useful in simplifying complex data spaces for machine learning tasks, however, to the best of our knowledge, they have not been sufficiently explored for regression problems. Granular balls group the data points into balls based on their proximity and reduce the computational cost in SVR by replacing the large number of data points with far fewer granular balls. This work also suggests a discretization method for continuous-valued attributes to facilitate the construction of granular balls. The effectiveness of the proposed approach is evaluated on several benchmark datasets and it outperforms existing state-of-the-art approaches
[ { "version": "v1", "created": "Thu, 13 Mar 2025 16:52:43 GMT" } ]
2025-03-14T00:00:00
[ [ "Rastogi", "Reshma", "" ], [ "Bisht", "Ankush", "" ], [ "Kumar", "Sanjay", "" ], [ "Chandra", "Suresh", "" ] ]
TITLE: GBSVR: Granular Ball Support Vector Regression ABSTRACT: Support Vector Regression (SVR) and its variants are widely used to handle regression tasks, however, since their solution involves solving an expensive quadratic programming problem, it limits its application, especially when dealing with large datasets. Additionally, SVR uses an epsilon-insensitive loss function which is sensitive to outliers and therefore can adversely affect its performance. We propose Granular Ball Support Vector Regression (GBSVR) to tackle problem of regression by using granular ball concept. These balls are useful in simplifying complex data spaces for machine learning tasks, however, to the best of our knowledge, they have not been sufficiently explored for regression problems. Granular balls group the data points into balls based on their proximity and reduce the computational cost in SVR by replacing the large number of data points with far fewer granular balls. This work also suggests a discretization method for continuous-valued attributes to facilitate the construction of granular balls. The effectiveness of the proposed approach is evaluated on several benchmark datasets and it outperforms existing state-of-the-art approaches
2503.10545
Vastal Srivastava Mr.
Vatsal Srivastava
From Linear to Spline-Based Classification:Developing and Enhancing SMPA for Noisy Non-Linear Datasets
null
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Building upon the concepts and mechanisms used for the development in Moving Points Algorithm, we will now explore how non linear decision boundaries can be developed for classification tasks. First we will look at the classification performance of MPA and some minor developments in the original algorithm. We then discuss the concepts behind using cubic splines for classification with a similar learning mechanism and finally analyze training results on synthetic datasets with known properties.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 16:58:40 GMT" } ]
2025-03-14T00:00:00
[ [ "Srivastava", "Vatsal", "" ] ]
TITLE: From Linear to Spline-Based Classification:Developing and Enhancing SMPA for Noisy Non-Linear Datasets ABSTRACT: Building upon the concepts and mechanisms used for the development in Moving Points Algorithm, we will now explore how non linear decision boundaries can be developed for classification tasks. First we will look at the classification performance of MPA and some minor developments in the original algorithm. We then discuss the concepts behind using cubic splines for classification with a similar learning mechanism and finally analyze training results on synthetic datasets with known properties.
2503.10560
Nicolas Pr\"ollochs
Kirill Solovev, Nicolas Pr\"ollochs
References to unbiased sources increase the helpfulness of community fact-checks
null
null
null
null
cs.SI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Community-based fact-checking is a promising approach to address misinformation on social media at scale. However, an understanding of what makes community-created fact-checks helpful to users is still in its infancy. In this paper, we analyze the determinants of the helpfulness of community-created fact-checks. For this purpose, we draw upon a unique dataset of real-world community-created fact-checks and helpfulness ratings from X's (formerly Twitter) Community Notes platform. Our empirical analysis implies that the key determinant of helpfulness in community-based fact-checking is whether users provide links to external sources to underpin their assertions. On average, the odds for community-created fact-checks to be perceived as helpful are 2.70 times higher if they provide links to external sources. Furthermore, we demonstrate that the helpfulness of community-created fact-checks varies depending on their level of political bias. Here, we find that community-created fact-checks linking to high-bias sources (of either political side) are perceived as significantly less helpful. This suggests that the rating mechanism on the Community Notes platform successfully penalizes one-sidedness and politically motivated reasoning. These findings have important implications for social media platforms, which can utilize our results to optimize their community-based fact-checking systems.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 17:12:01 GMT" } ]
2025-03-14T00:00:00
[ [ "Solovev", "Kirill", "" ], [ "Pröllochs", "Nicolas", "" ] ]
TITLE: References to unbiased sources increase the helpfulness of community fact-checks ABSTRACT: Community-based fact-checking is a promising approach to address misinformation on social media at scale. However, an understanding of what makes community-created fact-checks helpful to users is still in its infancy. In this paper, we analyze the determinants of the helpfulness of community-created fact-checks. For this purpose, we draw upon a unique dataset of real-world community-created fact-checks and helpfulness ratings from X's (formerly Twitter) Community Notes platform. Our empirical analysis implies that the key determinant of helpfulness in community-based fact-checking is whether users provide links to external sources to underpin their assertions. On average, the odds for community-created fact-checks to be perceived as helpful are 2.70 times higher if they provide links to external sources. Furthermore, we demonstrate that the helpfulness of community-created fact-checks varies depending on their level of political bias. Here, we find that community-created fact-checks linking to high-bias sources (of either political side) are perceived as significantly less helpful. This suggests that the rating mechanism on the Community Notes platform successfully penalizes one-sidedness and politically motivated reasoning. These findings have important implications for social media platforms, which can utilize our results to optimize their community-based fact-checking systems.
2503.10567
Nannan Wu
Nannan Wu, Zengqiang Yan, Nong Sang, Li Yu, Chang Wen Chen
FedPCA: Noise-Robust Fair Federated Learning via Performance-Capacity Analysis
Preprint
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Training a model that effectively handles both common and rare data-i.e., achieving performance fairness-is crucial in federated learning (FL). While existing fair FL methods have shown effectiveness, they remain vulnerable to mislabeled data. Ensuring robustness in fair FL is therefore essential. However, fairness and robustness inherently compete, which causes robust strategies to hinder fairness. In this paper, we attribute this competition to the homogeneity in loss patterns exhibited by rare and mislabeled data clients, preventing existing loss-based fair and robust FL methods from effectively distinguishing and handling these two distinct client types. To address this, we propose performance-capacity analysis, which jointly considers model performance on each client and its capacity to handle the dataset, measured by loss and a newly introduced feature dispersion score. This allows mislabeled clients to be identified by their significantly deviated performance relative to capacity while preserving rare data clients. Building on this, we introduce FedPCA, an FL method that robustly achieves fairness. FedPCA first identifies mislabeled clients via a Gaussian Mixture Model on loss-dispersion pairs, then applies fairness and robustness strategies in global aggregation and local training by adjusting client weights and selectively using reliable data. Extensive experiments on three datasets demonstrate FedPCA's effectiveness in tackling this complex challenge. Code will be publicly available upon acceptance.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 17:18:18 GMT" } ]
2025-03-14T00:00:00
[ [ "Wu", "Nannan", "" ], [ "Yan", "Zengqiang", "" ], [ "Sang", "Nong", "" ], [ "Yu", "Li", "" ], [ "Chen", "Chang Wen", "" ] ]
TITLE: FedPCA: Noise-Robust Fair Federated Learning via Performance-Capacity Analysis ABSTRACT: Training a model that effectively handles both common and rare data-i.e., achieving performance fairness-is crucial in federated learning (FL). While existing fair FL methods have shown effectiveness, they remain vulnerable to mislabeled data. Ensuring robustness in fair FL is therefore essential. However, fairness and robustness inherently compete, which causes robust strategies to hinder fairness. In this paper, we attribute this competition to the homogeneity in loss patterns exhibited by rare and mislabeled data clients, preventing existing loss-based fair and robust FL methods from effectively distinguishing and handling these two distinct client types. To address this, we propose performance-capacity analysis, which jointly considers model performance on each client and its capacity to handle the dataset, measured by loss and a newly introduced feature dispersion score. This allows mislabeled clients to be identified by their significantly deviated performance relative to capacity while preserving rare data clients. Building on this, we introduce FedPCA, an FL method that robustly achieves fairness. FedPCA first identifies mislabeled clients via a Gaussian Mixture Model on loss-dispersion pairs, then applies fairness and robustness strategies in global aggregation and local training by adjusting client weights and selectively using reliable data. Extensive experiments on three datasets demonstrate FedPCA's effectiveness in tackling this complex challenge. Code will be publicly available upon acceptance.
2503.10573
Afrar Jahin
Afrar Jahin, Arif Hassan Zidan, Yu Bao, Shizhe Liang, Tianming Liu, Wei Zhang
Unveiling the Mathematical Reasoning in DeepSeek Models: A Comparative Study of Large Language Models
null
null
null
null
cs.LG
http://creativecommons.org/licenses/by/4.0/
With the rapid evolution of Artificial Intelligence (AI), Large Language Models (LLMs) have reshaped the frontiers of various fields, spanning healthcare, public health, engineering, science, agriculture, education, arts, humanities, and mathematical reasoning. Among these advancements, DeepSeek models have emerged as noteworthy contenders, demonstrating promising capabilities that set them apart from their peers. While previous studies have conducted comparative analyses of LLMs, few have delivered a comprehensive evaluation of mathematical reasoning across a broad spectrum of LLMs. In this work, we aim to bridge this gap by conducting an in-depth comparative study, focusing on the strengths and limitations of DeepSeek models in relation to their leading counterparts. In particular, our study systematically evaluates the mathematical reasoning performance of two DeepSeek models alongside five prominent LLMs across three independent benchmark datasets. The findings reveal several key insights: 1). DeepSeek-R1 consistently achieved the highest accuracy on two of the three datasets, demonstrating strong mathematical reasoning capabilities. 2). The distilled variant of LLMs significantly underperformed compared to its peers, highlighting potential drawbacks in using distillation techniques. 3). In terms of response time, Gemini 2.0 Flash demonstrated the fastest processing speed, outperforming other models in efficiency, which is a crucial factor for real-time applications. Beyond these quantitative assessments, we delve into how architecture, training, and optimization impact LLMs' mathematical reasoning. Moreover, our study goes beyond mere performance comparison by identifying key areas for future advancements in LLM-driven mathematical reasoning. This research enhances our understanding of LLMs' mathematical reasoning and lays the groundwork for future advancements
[ { "version": "v1", "created": "Thu, 13 Mar 2025 17:23:45 GMT" } ]
2025-03-14T00:00:00
[ [ "Jahin", "Afrar", "" ], [ "Zidan", "Arif Hassan", "" ], [ "Bao", "Yu", "" ], [ "Liang", "Shizhe", "" ], [ "Liu", "Tianming", "" ], [ "Zhang", "Wei", "" ] ]
TITLE: Unveiling the Mathematical Reasoning in DeepSeek Models: A Comparative Study of Large Language Models ABSTRACT: With the rapid evolution of Artificial Intelligence (AI), Large Language Models (LLMs) have reshaped the frontiers of various fields, spanning healthcare, public health, engineering, science, agriculture, education, arts, humanities, and mathematical reasoning. Among these advancements, DeepSeek models have emerged as noteworthy contenders, demonstrating promising capabilities that set them apart from their peers. While previous studies have conducted comparative analyses of LLMs, few have delivered a comprehensive evaluation of mathematical reasoning across a broad spectrum of LLMs. In this work, we aim to bridge this gap by conducting an in-depth comparative study, focusing on the strengths and limitations of DeepSeek models in relation to their leading counterparts. In particular, our study systematically evaluates the mathematical reasoning performance of two DeepSeek models alongside five prominent LLMs across three independent benchmark datasets. The findings reveal several key insights: 1). DeepSeek-R1 consistently achieved the highest accuracy on two of the three datasets, demonstrating strong mathematical reasoning capabilities. 2). The distilled variant of LLMs significantly underperformed compared to its peers, highlighting potential drawbacks in using distillation techniques. 3). In terms of response time, Gemini 2.0 Flash demonstrated the fastest processing speed, outperforming other models in efficiency, which is a crucial factor for real-time applications. Beyond these quantitative assessments, we delve into how architecture, training, and optimization impact LLMs' mathematical reasoning. Moreover, our study goes beyond mere performance comparison by identifying key areas for future advancements in LLM-driven mathematical reasoning. This research enhances our understanding of LLMs' mathematical reasoning and lays the groundwork for future advancements
2503.10592
Hao He
Hao He, Ceyuan Yang, Shanchuan Lin, Yinghao Xu, Meng Wei, Liangke Gui, Qi Zhao, Gordon Wetzstein, Lu Jiang, Hongsheng Li
CameraCtrl II: Dynamic Scene Exploration via Camera-controlled Video Diffusion Models
Project page: https://hehao13.github.io/Projects-CameraCtrl-II/
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
This paper introduces CameraCtrl II, a framework that enables large-scale dynamic scene exploration through a camera-controlled video diffusion model. Previous camera-conditioned video generative models suffer from diminished video dynamics and limited range of viewpoints when generating videos with large camera movement. We take an approach that progressively expands the generation of dynamic scenes -- first enhancing dynamic content within individual video clip, then extending this capability to create seamless explorations across broad viewpoint ranges. Specifically, we construct a dataset featuring a large degree of dynamics with camera parameter annotations for training while designing a lightweight camera injection module and training scheme to preserve dynamics of the pretrained models. Building on these improved single-clip techniques, we enable extended scene exploration by allowing users to iteratively specify camera trajectories for generating coherent video sequences. Experiments across diverse scenarios demonstrate that CameraCtrl Ii enables camera-controlled dynamic scene synthesis with substantially wider spatial exploration than previous approaches.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 17:42:01 GMT" } ]
2025-03-14T00:00:00
[ [ "He", "Hao", "" ], [ "Yang", "Ceyuan", "" ], [ "Lin", "Shanchuan", "" ], [ "Xu", "Yinghao", "" ], [ "Wei", "Meng", "" ], [ "Gui", "Liangke", "" ], [ "Zhao", "Qi", "" ], [ "Wetzstein", "Gordon", ...
TITLE: CameraCtrl II: Dynamic Scene Exploration via Camera-controlled Video Diffusion Models ABSTRACT: This paper introduces CameraCtrl II, a framework that enables large-scale dynamic scene exploration through a camera-controlled video diffusion model. Previous camera-conditioned video generative models suffer from diminished video dynamics and limited range of viewpoints when generating videos with large camera movement. We take an approach that progressively expands the generation of dynamic scenes -- first enhancing dynamic content within individual video clip, then extending this capability to create seamless explorations across broad viewpoint ranges. Specifically, we construct a dataset featuring a large degree of dynamics with camera parameter annotations for training while designing a lightweight camera injection module and training scheme to preserve dynamics of the pretrained models. Building on these improved single-clip techniques, we enable extended scene exploration by allowing users to iteratively specify camera trajectories for generating coherent video sequences. Experiments across diverse scenarios demonstrate that CameraCtrl Ii enables camera-controlled dynamic scene synthesis with substantially wider spatial exploration than previous approaches.
2503.10596
Tianheng Cheng
Rui Hu, Lianghui Zhu, Yuxuan Zhang, Tianheng Cheng, Lei Liu, Heng Liu, Longjin Ran, Xiaoxin Chen, Wenyu Liu, Xinggang Wang
GroundingSuite: Measuring Complex Multi-Granular Pixel Grounding
Work in progress. Code: https://github.com/hustvl/GroundingSuite
null
null
null
cs.CV
http://creativecommons.org/licenses/by-nc-sa/4.0/
Pixel grounding, encompassing tasks such as Referring Expression Segmentation (RES), has garnered considerable attention due to its immense potential for bridging the gap between vision and language modalities. However, advancements in this domain are currently constrained by limitations inherent in existing datasets, including limited object categories, insufficient textual diversity, and a scarcity of high-quality annotations. To mitigate these limitations, we introduce GroundingSuite, which comprises: (1) an automated data annotation framework leveraging multiple Vision-Language Model (VLM) agents; (2) a large-scale training dataset encompassing 9.56 million diverse referring expressions and their corresponding segmentations; and (3) a meticulously curated evaluation benchmark consisting of 3,800 images. The GroundingSuite training dataset facilitates substantial performance improvements, enabling models trained on it to achieve state-of-the-art results. Specifically, a cIoU of 68.9 on gRefCOCO and a gIoU of 55.3 on RefCOCOm. Moreover, the GroundingSuite annotation framework demonstrates superior efficiency compared to the current leading data annotation method, i.e., $4.5 \times$ faster than the GLaMM.
[ { "version": "v1", "created": "Thu, 13 Mar 2025 17:43:10 GMT" } ]
2025-03-14T00:00:00
[ [ "Hu", "Rui", "" ], [ "Zhu", "Lianghui", "" ], [ "Zhang", "Yuxuan", "" ], [ "Cheng", "Tianheng", "" ], [ "Liu", "Lei", "" ], [ "Liu", "Heng", "" ], [ "Ran", "Longjin", "" ], [ "Chen", "Xiaoxin", ...
TITLE: GroundingSuite: Measuring Complex Multi-Granular Pixel Grounding ABSTRACT: Pixel grounding, encompassing tasks such as Referring Expression Segmentation (RES), has garnered considerable attention due to its immense potential for bridging the gap between vision and language modalities. However, advancements in this domain are currently constrained by limitations inherent in existing datasets, including limited object categories, insufficient textual diversity, and a scarcity of high-quality annotations. To mitigate these limitations, we introduce GroundingSuite, which comprises: (1) an automated data annotation framework leveraging multiple Vision-Language Model (VLM) agents; (2) a large-scale training dataset encompassing 9.56 million diverse referring expressions and their corresponding segmentations; and (3) a meticulously curated evaluation benchmark consisting of 3,800 images. The GroundingSuite training dataset facilitates substantial performance improvements, enabling models trained on it to achieve state-of-the-art results. Specifically, a cIoU of 68.9 on gRefCOCO and a gIoU of 55.3 on RefCOCOm. Moreover, the GroundingSuite annotation framework demonstrates superior efficiency compared to the current leading data annotation method, i.e., $4.5 \times$ faster than the GLaMM.