Every paper checked — nothing hidden

All papers

Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.

All17,312 In brief24 More (on-topic)3,333 Not relevant13,753 Not assessed87 Older351 Could not confirm93 Blogs with no paper266 All blogs284

Paper	Topic	Relevance
Purify Before You Align: Improving Representation Alignment for Diffusion Models Introduces local appearance purification to filter alignment targets and improve representation learning during diffusion model training. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Focuses on image diffusion training targets rather than multimodal integration.
Purify Before You Align: Improving Representation Alignment for Diffusion Models Introduces local appearance purification to filter alignment targets and improve representation learning during diffusion model training. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:18	Not relevant Studies diffusion representation alignment, not AI safety or deployment guardrails.
Blame Errant Words, Not the Canvas: Defect-Aware Spatial Credit Assignment for Visual Text Generation Provides region-specific reward modeling to fix incorrect wording and glyph errors in visual text generation models. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Studies text rendering in generative images, not industrial inspection.
RIGen: Beyond Action Following with Reciprocal Interaction in Generative Gameplay Combines state-tracking rules with action-conditioned video generation to enable persistent reciprocal interactions in generative gameplay. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Covers generative video gameplay, not industrial inspection or defect detection.
Intern Lumina U2: A Multi-Codebook Diffusion Large Language Model for Omni-Visual Understanding and Image Generation Uses a multi-codebook diffusion language model to jointly represent text and discrete visual tokens for multimodal generation. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Focuses on generative image-text modeling, not industrial defect detection.
MARS - A foundational map auto-regressor Introduces an autoregressive foundation model to generate vectorized road networks and building polygons directly from spatial inputs. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Deals with map extraction from overhead imagery, not industrial quality inspection.
DefoliationCAM: An Edge-AI System for Real-Time Insect Defoliation Severity Assessment in Grapevines Builds a solar-powered edge vision system to detect and quantify insect defoliation severity in grapevine canopies. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Applies vision to agricultural pest defoliation, not industrial defect detection.
DefoliationCAM: An Edge-AI System for Real-Time Insect Defoliation Severity Assessment in Grapevines Builds a solar-powered edge vision system to detect and quantify insect defoliation severity in grapevine canopies. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Applies standard computer vision pipelines rather than multimodal vision-language reasoning.
Pixels to Tokens: Token Space Efficient Active Learning for Low-Budget Semantic Segmentation Selects maximally informative pixel tokens directly from vision transformers to train segmentation heads under tight annotation budgets. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant General semantic segmentation active learning, not industrial manufacturing defect detection.
Pixels to Tokens: Token Space Efficient Active Learning for Low-Budget Semantic Segmentation Selects maximally informative pixel tokens directly from vision transformers to train segmentation heads under tight annotation budgets. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Focuses solely on computer vision ViT segmentation rather than multimodal models.
Pixels to Tokens: Token Space Efficient Active Learning for Low-Budget Semantic Segmentation Selects maximally informative pixel tokens directly from vision transformers to train segmentation heads under tight annotation budgets. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:18	Not relevant Focuses on active learning segmentation rather than AI safety or reliability.
Pixels to Tokens: Token Space Efficient Active Learning for Low-Budget Semantic Segmentation Selects maximally informative pixel tokens directly from vision transformers to train segmentation heads under tight annotation budgets. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18	Not relevant Addresses active learning algorithms, not LLM evaluation or benchmarks.
Pixels to Tokens: Token Space Efficient Active Learning for Low-Budget Semantic Segmentation Selects maximally informative pixel tokens directly from vision transformers to train segmentation heads under tight annotation budgets. Preprint	Small and efficient models via arxiv · 17 Oct, 06:18	Not relevant Covers active learning annotation budgets rather than model compression or efficient inference.
EgoJEPA: Beyond Retargeting for Dexterous Manipulation from Egocentric Videos Learns robot manipulation priors from egocentric human videos using action prediction rather than rule-based hand retargeting. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Focuses on visual robot control rather than multimodal language or document understanding.
EgoJEPA: Beyond Retargeting for Dexterous Manipulation from Egocentric Videos Learns robot manipulation priors from egocentric human videos using action prediction rather than rule-based hand retargeting. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Focuses on robotic manipulation from egocentric video, not industrial defect detection.
Decision-FLIP: Budget-Free Structured Pruning for Gaussian Occupancy Models Prunes Gaussian occupancy models without preset budgets by identifying and removing Gaussians that do not alter voxel classification decisions. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Covers 3D occupancy pruning, not manufacturing defect detection or visual quality control.
Decision-FLIP: Budget-Free Structured Pruning for Gaussian Occupancy Models Prunes Gaussian occupancy models without preset budgets by identifying and removing Gaussians that do not alter voxel classification decisions. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Addresses 3D occupancy model pruning rather than vision-language or multimodal cross-modal modeling.
Line-Augmented cuVSLAM: Robust Visual Odometry in Low-Texture Environments Adds line-feature tracking to GPU-accelerated visual odometry to improve warehouse robot localization in low-texture environments. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Focuses on mobile robot visual odometry, not industrial defect detection or visual quality control.
Line-Augmented cuVSLAM: Robust Visual Odometry in Low-Texture Environments Adds line-feature tracking to GPU-accelerated visual odometry to improve warehouse robot localization in low-texture environments. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Addresses robotics visual SLAM rather than multimodal vision-language models.
PiCo: Physics-Informed Concept Probing for Geo-Foundation Models Uses synthetic physics-based satellite observations to probe and explain physical feature sensitivity inside geo-foundation model representations. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Focuses on remote sensing satellite models, not industrial manufacturing inspection.
Do As I Do, Not As I Say: On-Policy Distillation Inherits the Teacher in Vision-Language Models Evaluates on-policy distillation versus supervised fine-tuning in vision-language models, analyzing how teacher reliability impacts student behavior. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:18	Not relevant Focuses on post-training distillation accuracy, not safety or guardrails.
Do As I Do, Not As I Say: On-Policy Distillation Inherits the Teacher in Vision-Language Models Evaluates on-policy distillation versus supervised fine-tuning in vision-language models, analyzing how teacher reliability impacts student behavior. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Studies distillation on vision-language reasoning, not manufacturing defect inspection.
Do As I Do, Not As I Say: On-Policy Distillation Inherits the Teacher in Vision-Language Models Evaluates on-policy distillation versus supervised fine-tuning in vision-language models, analyzing how teacher reliability impacts student behavior. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18	Not relevant Investigates distillation training behavior rather than LLM benchmark design.
DeSpatial: The devil of Agentic Spatial Reasoning Lies in 3D Itself Decomposes agentic spatial reasoning by using a metric 3D perception tool to resolve visual inconsistencies before spatial reasoning. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Addresses spatial agent reasoning benchmarks, not industrial quality control.
Mind the Tail: Data Selection Through Sparse Concepts Uses sparse autoencoders to select diverse training samples and prevent underrepresenting tail concepts across vision and multimodal models. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Focuses on dataset curation for general image and multimodal pretraining.
Audio-Visual Dog Identification and Tracking in Real-World Videos Combines visual tracking and vocal fingerprinting to identify and monitor individual dogs across real-world multi-dog video recordings. Preprint	Small and efficient models via arxiv · 17 Oct, 06:18	Not relevant Focuses on multi-modal tracking, not model compression or serving.
Audio-Visual Dog Identification and Tracking in Real-World Videos Combines visual tracking and vocal fingerprinting to identify and monitor individual dogs across real-world multi-dog video recordings. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18	Not relevant Introduces an animal audio-visual tracking benchmark, not LLM evaluation.
Audio-Visual Dog Identification and Tracking in Real-World Videos Combines visual tracking and vocal fingerprinting to identify and monitor individual dogs across real-world multi-dog video recordings. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:18	Not relevant Covers animal tracking, not deployment safety or guardrails.
Audio-Visual Dog Identification and Tracking in Real-World Videos Combines visual tracking and vocal fingerprinting to identify and monitor individual dogs across real-world multi-dog video recordings. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Studies animal tracking from video and audio, not industrial inspection.
RoPE-Flow: Monocular Vision-based Robot Pose Estimation via 2D Graph-Conditioned 3D Flow Matching Estimates 3D robot pose from monocular images using graph-guided 2D keypoints and conditional flow matching. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Studies monocular 3D robot pose estimation without multimodal integration.
RoPE-Flow: Monocular Vision-based Robot Pose Estimation via 2D Graph-Conditioned 3D Flow Matching Estimates 3D robot pose from monocular images using graph-guided 2D keypoints and conditional flow matching. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Covers monocular robot pose estimation, not manufacturing defect detection.
RoboCloth: A Large-Scale Real Cloth Material Dataset for Neural Reflectance Reconstruction Presents a dataset of 500 robotically imaged cloth materials for training and evaluating neural reflectance reconstruction models. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Studies neural cloth reflectance reconstruction, not industrial visual inspection.
RoboCloth: A Large-Scale Real Cloth Material Dataset for Neural Reflectance Reconstruction Presents a dataset of 500 robotically imaged cloth materials for training and evaluating neural reflectance reconstruction models. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Focuses on 3D neural reflectance modeling, not multimodal models.
Can World and Action Modeling Benefit Each Other under Unified Causal Training? Tests whether jointly training action-conditioned world models and robot policy models in a single architecture improves both tasks. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Focuses on robot manipulation dynamics, not industrial defect detection.
Spectral-Spatial Interpretation Introduces a Shapley-based attribution method that jointly explains spatial and spectral channel importance in multispectral imagery. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18	Not relevant Presents an attribution interpretability technique, not LLM benchmark evaluation.
Spectral-Spatial Interpretation Introduces a Shapley-based attribution method that jointly explains spatial and spectral channel importance in multispectral imagery. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Explores interpretability for multispectral earth observation, not factory inspection.
Spectral-Spatial Interpretation Introduces a Shapley-based attribution method that jointly explains spatial and spectral channel importance in multispectral imagery. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Focuses on multispectral image interpretability rather than multimodal LLMs.
Spectral-Spatial Interpretation Introduces a Shapley-based attribution method that jointly explains spatial and spectral channel importance in multispectral imagery. Preprint	Small and efficient models via arxiv · 17 Oct, 06:18	Not relevant Focuses on feature attribution, not model compression or efficiency.
Spectral-Spatial Interpretation Introduces a Shapley-based attribution method that jointly explains spatial and spectral channel importance in multispectral imagery. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:18	Not relevant Studies post-hoc feature attribution rather than AI safety guardrails.
ResStreamGS: Residual-Sensitive Dual-Stream Gaussian Splatting for Mobile Volumetric Video Streaming Separates static and dynamic 3D Gaussians to enable efficient real-time volumetric video streaming on mobile devices. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Addresses volumetric video streaming, not factory or defect inspection.
ResStreamGS: Residual-Sensitive Dual-Stream Gaussian Splatting for Mobile Volumetric Video Streaming Separates static and dynamic 3D Gaussians to enable efficient real-time volumetric video streaming on mobile devices. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Covers 3D Gaussian Splatting video streaming, not multimodal models.
Sparse Attention for Video Generation Acceleration via Growing Sparsity and Reduced Search Accelerates diffusion transformer video generation by increasing attention sparsity and reusing search indices across denoising steps. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Focuses on video generation acceleration, not industrial visual inspection.
Sparse Attention for Video Generation Acceleration via Growing Sparsity and Reduced Search Accelerates diffusion transformer video generation by increasing attention sparsity and reusing search indices across denoising steps. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Focuses on accelerating video generation diffusion transformers without multimodal reasoning.
Trajectory Prediction Method for person under WSN with Hierarchical Envelope Adversarial Domain Adaptation Develops a domain adaptation framework to improve pedestrian trajectory prediction under varied social couplings and scene shifts. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Studies trajectory prediction rather than multimodal language or vision models.
Trajectory Prediction Method for person under WSN with Hierarchical Envelope Adversarial Domain Adaptation Develops a domain adaptation framework to improve pedestrian trajectory prediction under varied social couplings and scene shifts. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Focuses on pedestrian trajectory prediction, not manufacturing defect detection.
VisualTrust: Video Reasoning under On-Screen Claim Framing Introduces a counterfactual benchmark to evaluate whether multimodal models erroneously defer to misleading on-screen text over video evidence. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Analyzes video reasoning under misleading text, not industrial inspection.
Monotone Piecewise-Linear Harmonization of Multitemporal Multispectral Surface Reflectance With Unsupervised Deep Multiview Pseudoinvariant Features Learning Presents an unsupervised deep multiview learning framework to correct systematic multitemporal surface reflectance inconsistencies across satellite observations. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Concerns satellite remote sensing reflectance, not industrial vision applications.
Monotone Piecewise-Linear Harmonization of Multitemporal Multispectral Surface Reflectance With Unsupervised Deep Multiview Pseudoinvariant Features Learning Presents an unsupervised deep multiview learning framework to correct systematic multitemporal surface reflectance inconsistencies across satellite observations. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Focuses on multispectral satellite imagery rather than multimodal language-vision models.
ControlFlow3D: Distilling Multi-View Knowledge into Latent Flow Matching for Point Cloud Upsampling Distills multi-view visual cues into a latent flow matching network to improve point cloud upsampling at inference time. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Deals with 3D point cloud upsampling, not manufacturing inspection.
UltraDiff: Transferring High-Fidelity Priors to Compressed Latent Spaces for High-Resolution Image Generation Distills high-fidelity priors into heavily compressed latent spaces to enable faster high-resolution image generation. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Studies generative diffusion models, not industrial quality control.
UltraDiff: Transferring High-Fidelity Priors to Compressed Latent Spaces for High-Resolution Image Generation Distills high-fidelity priors into heavily compressed latent spaces to enable faster high-resolution image generation. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Focuses on image generation efficiency rather than multimodal reasoning.
SpaceCache: Intermediate Feature Reuse for Efficient Semantic Segmentation on Point Cloud Streams Reuses intermediate transformer features across video frames to speed up 3D point cloud semantic segmentation without extra training. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Deals with autonomous driving point clouds, not industrial inspection.
SpaceCache: Intermediate Feature Reuse for Efficient Semantic Segmentation on Point Cloud Streams Reuses intermediate transformer features across video frames to speed up 3D point cloud semantic segmentation without extra training. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Studies unimodal 3D point cloud streaming feature reuse.
DLOSAC: Joint Spatial-Geometric Image Estimation via Delaunay-Guided Local Optimization RANSAC Proposes a local optimization method using Delaunay triangulation to balance geometric consistency with spatial distribution in RANSAC estimation. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Presents generic RANSAC geometric matching, not industrial defect detection.
DLOSAC: Joint Spatial-Geometric Image Estimation via Delaunay-Guided Local Optimization RANSAC Proposes a local optimization method using Delaunay triangulation to balance geometric consistency with spatial distribution in RANSAC estimation. Preprint	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Focuses on unimodal geometric image estimation without multimodal data.
A large-scale climate-aware satellite image dataset for domain adaptive land-cover semantic segmentation Introduces a climate-aware satellite image dataset to evaluate and improve domain adaptive land-cover semantic segmentation across different regions. Peer-reviewed	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Covers satellite remote sensing land-cover segmentation, not industrial factory inspection.
A large-scale climate-aware satellite image dataset for domain adaptive land-cover semantic segmentation Introduces a climate-aware satellite image dataset to evaluate and improve domain adaptive land-cover semantic segmentation across different regions. Peer-reviewed	Multimodal models via arxiv · 17 Oct, 06:18	Not relevant Deals solely with unimodal satellite image semantic segmentation.
FocusOPD: Structure-Guided On-Policy Distillation for Diffusion Models Improves on-policy distillation for diffusion models by restricting feature alignment supervision to deep layers and early denoising steps. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Focuses on generative diffusion models, not industrial visual inspection.
Do VLAs Really Follow Instructions? Diagnosing How-Level Instruction Following in Vision-Language-Action Models Studies how vision-language-action models fail to follow process instructions and introduces a benchmark to measure instruction compliance. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Focuses on embodied robotic instruction following, not industrial defect detection.
LICOA: Learning Instruction-Conditioned Optical Actions for Mobile Spatial Augmented Reality Introduces a robotic system that converts language instructions and visual inputs into physical light projection actions in spatial environments. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:18	Not relevant Focuses on mobile spatial augmented reality, not industrial visual inspection.

601–660 of 13,753← Previous Next →