Every paper checked — nothing hidden

All papers

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

All17,313 In brief24 More (on-topic)3,333 Not relevant13,754 Not assessed87 Older352 Could not confirm93 Blogs with no paper266 All blogs284

Paper	Topic	Relevance
Multi-modal Representation Fusion for Long-term Microsurgical Anastomosis Trajectory Anticipation This paper proposes a cross-modal representation fusion method to predict long-term tool trajectories during microsurgical anastomosis procedures. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Fuses surgical video and motion trajectories, not multimodal vision-language models.
Multi-modal Representation Fusion for Long-term Microsurgical Anastomosis Trajectory Anticipation This paper proposes a cross-modal representation fusion method to predict long-term tool trajectories during microsurgical anastomosis procedures. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Microsurgical trajectory anticipation, not industrial inspection or manufacturing quality control.
DSR-Net: Learning Dynamic Skeletal Relations for 3D Human Pose Estimation This paper proposes a framework that learns dynamic skeletal joint relationships and temporal interactions for 3D human pose estimation. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Human 3D pose estimation, not industrial inspection or defect detection.
DSR-Net: Learning Dynamic Skeletal Relations for 3D Human Pose Estimation This paper proposes a framework that learns dynamic skeletal joint relationships and temporal interactions for 3D human pose estimation. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Focuses on 3D skeletal pose estimation from vision, not multimodal models.
Spatiotemporal Recognition of Abdominal Wall Layers During Optical Trocar Entry This paper evaluates spatial and temporal modeling methods to improve multi-class tissue recognition during surgical optical trocar insertion. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Applies spatiotemporal computer vision to medical videos, not multimodal models.
Spatiotemporal Recognition of Abdominal Wall Layers During Optical Trocar Entry This paper evaluates spatial and temporal modeling methods to improve multi-class tissue recognition during surgical optical trocar insertion. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Surgical tissue recognition, not manufacturing or industrial quality control.
TartanMatch: Towards Universal Dense Correspondence Across Modalities This paper presents a unified model for dense visual correspondence across 25 pairs of multi-modal sensor inputs. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Universal robotic sensor correspondence, not industrial visual inspection or defect detection.
TartanMatch: Towards Universal Dense Correspondence Across Modalities This paper presents a unified model for dense visual correspondence across 25 pairs of multi-modal sensor inputs. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant A cross-sensor matching model, not a small language model or model compression technique.
TartanMatch: Towards Universal Dense Correspondence Across Modalities This paper presents a unified model for dense visual correspondence across 25 pairs of multi-modal sensor inputs. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on cross-modal correspondence, not AI safety, alignment, or guardrails.
TartanMatch: Towards Universal Dense Correspondence Across Modalities This paper presents a unified model for dense visual correspondence across 25 pairs of multi-modal sensor inputs. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Proposes a cross-modal correspondence model, not an LLM evaluation benchmark.
We Have Been Interpreting Diffusion Transformer Backbone Behaviors Wrong This paper analyzes intermediate layer activations in diffusion transformer backbones to explain why layer performance follows an inverted-U curve. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Analyzes internal layer representations of diffusion vision models, not multimodal models.
We Have Been Interpreting Diffusion Transformer Backbone Behaviors Wrong This paper analyzes intermediate layer activations in diffusion transformer backbones to explain why layer performance follows an inverted-U curve. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Diffusion transformer analysis, not industrial visual inspection or defect detection.
InfraWorld: Motion-Aware World Modeling for Roadside 4D Occupancy Forecasting Proposes a world model combining dense bird's-eye-view representations and motion queries for roadside dynamic occupancy forecasting. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Autonomous driving roadside occupancy forecasting, not manufacturing defect detection.
InfraWorld: Motion-Aware World Modeling for Roadside 4D Occupancy Forecasting Proposes a world model combining dense bird's-eye-view representations and motion queries for roadside dynamic occupancy forecasting. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Deals with 3D/4D spatial occupancy forecasting, not multimodal vision-language models.
Adversarial Gradient Forcing for Robust Few-Step Autoregressive Video Generation Introduces an adversarial training method to make few-step autoregressive video models robust against accumulated rollout errors. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Deals with generative video modeling, not multimodal models.
Adversarial Gradient Forcing for Robust Few-Step Autoregressive Video Generation Introduces an adversarial training method to make few-step autoregressive video models robust against accumulated rollout errors. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Autoregressive video generation robustness, not industrial visual inspection or defect detection.
FrameFold: Response-Guided KV Cache Consolidation for Autoregressive Long Video Generation Proposes a training-free KV cache consolidation technique guided by historical queries to manage memory in long video generation. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Video generation KV cache consolidation, not industrial visual inspection or defect detection.
FrameFold: Response-Guided KV Cache Consolidation for Autoregressive Long Video Generation Proposes a training-free KV cache consolidation technique guided by historical queries to manage memory in long video generation. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Focuses purely on video generation KV cache optimization, not cross-modal understanding.
Remembering What Matters for Long-Horizon Robotic World Models Develops a context-scheduling method that preserves anchor frames and evicts stale history in long-horizon robotic world models. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Focuses on robotic video world models and context scheduling, not multimodal reasoning.
Remembering What Matters for Long-Horizon Robotic World Models Develops a context-scheduling method that preserves anchor frames and evicts stale history in long-horizon robotic world models. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Robotic world models, not industrial visual inspection or defect detection.
Flow-Support-Weighted Proposals for Test-Time Verified VLA Policies Introduces a training-time sampling method to prevent action proposal collapse in vision-language-action models tested with verifiers. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Robotic policy action sampling, not industrial visual inspection or defect detection.
Attribute-Centric Vision-Language Alignment for Fine-Grained Gait Understanding Presents a dataset and vision-language alignment framework for fine-grained human gait attribute recognition from walking sequences. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Human gait analysis, not industrial visual inspection or defect detection.
ToF-Wheel Odometry: Navigating Degeneration of Localization With Limited Sensing Capabilities Combines wheel kinematics with Time-of-Flight camera geometry using directional covariance shaping to prevent odometry divergence. Peer-reviewed	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Robot odometry and localization, not industrial defect detection or visual quality control.
ToF-Wheel Odometry: Navigating Degeneration of Localization With Limited Sensing Capabilities Combines wheel kinematics with Time-of-Flight camera geometry using directional covariance shaping to prevent odometry divergence. Peer-reviewed	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Addresses robot state estimation from wheel odometry and ToF sensors, not multimodal models.
EviScene: Beyond "Is It AI?" for Agentic Image Factuality Assessment Introduces a benchmark and multimodal agent that uses visual inspection and external retrieval to verify real-world image factuality. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Image factuality verification, not industrial visual inspection or defect detection.
DuSRFlow: Dual-Lens Super-Resolution via Parallax-Aware Flow Estimation and Complementary Reference Warping Develops a dedicated flow estimator and reference warping method for dual-lens super-resolution under large parallax conditions. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Consumer camera super-resolution, not industrial visual inspection or defect detection.
DuSRFlow: Dual-Lens Super-Resolution via Parallax-Aware Flow Estimation and Complementary Reference Warping Develops a dedicated flow estimator and reference warping method for dual-lens super-resolution under large parallax conditions. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Focuses purely on dual-lens image super-resolution, not multimodal models.
Automatic assembly modeling through attention-based assembly feature mate prediction and assembly feature pattern recognition Presents an attention-based assembly feature mate prediction model to automate CAD assembly modeling from unassembled boundary representation parts. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Deals with 3D CAD B-Rep geometry, not multimodal vision-language models.
Automatic assembly modeling through attention-based assembly feature mate prediction and assembly feature pattern recognition Presents an attention-based assembly feature mate prediction model to automate CAD assembly modeling from unassembled boundary representation parts. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant CAD assembly modeling, not manufacturing defect detection or visual quality control.
Can Synthetic UAV Benchmarks Be Trusted? Real-Synthetic Validation for UAV Model Selection Investigates whether synthetic aerial imagery benchmarks preserve model-selection rankings compared to evaluations on real drone data. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Synthetic UAV imagery evaluation, not industrial manufacturing quality control.
Can Synthetic UAV Benchmarks Be Trusted? Real-Synthetic Validation for UAV Model Selection Investigates whether synthetic aerial imagery benchmarks preserve model-selection rankings compared to evaluations on real drone data. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant UAV aerial perception validation, not multimodal foundation model research.
TAG: Calibrating SAM3 for Universal Image Segmentation without Dense Supervision Proposes a text anchor re-grounding method using low-rank adapters to improve concept recognition in Segment Anything Model 3. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Universal image segmentation, not industrial defect detection.
TGuide: Faster Autoregressive Video Diffusion via Historical Trajectory Guidance Accelerates few-step autoregressive video diffusion by predicting features from historical sampling trajectories across video chunks. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Video diffusion acceleration, not manufacturing anomaly detection.
TGuide: Faster Autoregressive Video Diffusion via Historical Trajectory Guidance Accelerates few-step autoregressive video diffusion by predicting features from historical sampling trajectories across video chunks. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Video diffusion sampling acceleration, not multimodal models.
TETRIS-4DGS: Efficient Post-Training Compression of 4D Gaussian Splatting Introduces an adaptive quantization and packing framework to compress 4D Gaussian Splatting dynamic scenes using standard video codecs. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Compression for 4D Gaussian splats, not industrial inspection.
TETRIS-4DGS: Efficient Post-Training Compression of 4D Gaussian Splatting Introduces an adaptive quantization and packing framework to compress 4D Gaussian Splatting dynamic scenes using standard video codecs. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Dynamic scene rendering representation, not multimodal reasoning.
SPARE: Shared Perception and Adaptive Retrieval of Evidence for Streaming Video Understanding Presents an adaptive retrieval framework that selectively retrieves historical video context to assist streaming question answering. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Streaming video question answering, not industrial defect detection.
Competition Domains Matter: Selective Head Globalization for Training-Free Sparse Attention in Video Diffusion Transformers Develops a training-free sparse attention routing method that adapts competition domains across attention heads in video diffusion transformers. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Attention optimization for video generation, not multimodal model reasoning.
Competition Domains Matter: Selective Head Globalization for Training-Free Sparse Attention in Video Diffusion Transformers Develops a training-free sparse attention routing method that adapts competition domains across attention heads in video diffusion transformers. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Efficient attention for video diffusion, not industrial defect detection.
CAST-ASD: Causal Attention over Speakers and Time for Online Active Speaker Detection Proposes a causal joint attention model across faces and time for low-latency active speaker detection in multi-person video. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Active speaker detection, not industrial anomaly detection.
HumanPhys: A Human-Centric Physical-World Perception Benchmark for Real-World Visual Reasoning Introduces a benchmark probing physical perception tasks related to humans, including depth, trajectory, illumination, and kinematics. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Physical world human reasoning, not factory visual inspection.
Demiurge: Forging Navigable 4D Worlds via Autoregressive Video Rerendering Presents an autoregressive video rerendering pipeline using a dynamic 4D point-cloud memory cache for novel-view video synthesis. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Video generation and rerendering, not multimodal foundation models.
Demiurge: Forging Navigable 4D Worlds via Autoregressive Video Rerendering Presents an autoregressive video rerendering pipeline using a dynamic 4D point-cloud memory cache for novel-view video synthesis. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Novel-view video synthesis, not industrial defect inspection.
PURE: Refining Feed-forward 3D Reconstruction with Unlabeled Post-training Refines feed-forward 3D reconstruction models on dynamic video through unlabeled post-training using relative depth constraints and 2D priors. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Computer vision 3D reconstruction from video, not multimodal reasoning.
PURE: Refining Feed-forward 3D Reconstruction with Unlabeled Post-training Refines feed-forward 3D reconstruction models on dynamic video through unlabeled post-training using relative depth constraints and 2D priors. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Dynamic video 3D reconstruction, not manufacturing inspection.
PixARG: Pixel-Asynchronous Retrieval-Augmented Diffusion for Fast and Realistic Image Generation Integrates real-world image retrieval into diffusion models using pixel-level affinity maps and heterogeneous noise initialization to accelerate generation. Peer-reviewed	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Retrieval-augmented image synthesis, not industrial quality control.
LightVAE: Towards Compact and Efficient Video Autoencoders Proposes a video autoencoder compression framework that uses temporal low-rank approximation to reduce computation in latent video diffusion. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Video compression for diffusion models, not industrial inspection.
LightVAE: Towards Compact and Efficient Video Autoencoders Proposes a video autoencoder compression framework that uses temporal low-rank approximation to reduce computation in latent video diffusion. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Pure video autoencoder architecture, not multimodal vision-language learning.
Align-VAE: Aligning Attention and Geometry in VecSet VAEs for 3D Reconstruction Introduces modifications to set-based variational autoencoders to improve geometric alignment and attention quality in 3D shape reconstruction. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant 3D shape latent representation learning, not cross-modal understanding.
Align-VAE: Aligning Attention and Geometry in VecSet VAEs for 3D Reconstruction Introduces modifications to set-based variational autoencoders to improve geometric alignment and attention quality in 3D shape reconstruction. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant 3D geometric shape reconstruction, not industrial defect detection.
Japanese Handwriting Generation Based on Human Preference Feedback Using Direct Preference Optimization Applies Direct Preference Optimization to align generative models with human handwriting preferences and capture natural shape variability. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Image generation of handwriting styles, not multimodal vision-language understanding.
Japanese Handwriting Generation Based on Human Preference Feedback Using Direct Preference Optimization Applies Direct Preference Optimization to align generative models with human handwriting preferences and capture natural shape variability. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Handwriting generation, not industrial visual quality control.
BridgeWAM: Bridging Video and Action Experts via Latent Bridge Queries The paper presents an architecture that decouples video and action prediction models using latent bridge queries. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Robot manipulation action models, not industrial visual inspection.
[Re] Boosting the Visual Interpretability of CLIP via Adversarial Fine-Tuning The paper presents a reproducibility study and an explicit saliency supervision extension for adversarial fine-tuning of CLIP vision models. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Interpretable CLIP fine-tuning, not industrial inspection or quality control.
Towards Privacy-Preserving Thermal Human Perception from Dataset to Deployment The paper presents an indoor paired thermal and RGB dataset and benchmarks for privacy-preserving human occupancy detection. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Paired RGB-thermal vision perception, not general multimodal reasoning or VLMs.
Towards Privacy-Preserving Thermal Human Perception from Dataset to Deployment The paper presents an indoor paired thermal and RGB dataset and benchmarks for privacy-preserving human occupancy detection. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Indoor human perception and sleep monitoring, not industrial manufacturing.
CAST: Real-Time Motion Capture for Any Skeleton Topology The paper proposes a motion capture framework that predicts and aligns joint rotations across arbitrary target skeleton topologies from monocular video. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Single-modality 3D pose and motion capture from monocular video.
CAST: Real-Time Motion Capture for Any Skeleton Topology The paper proposes a motion capture framework that predicts and aligns joint rotations across arbitrary target skeleton topologies from monocular video. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Monocular motion capture, not industrial visual inspection or defect detection.
ScalLoc: Continuous Oblique Thermal UAV Geo-Localization with a 2D Satellite Map The paper introduces a framework for continuous oblique thermal drone geo-localization using 2D satellite maps in navigation-denied settings. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant UAV geo-localization, not industrial defect detection or manufacturing inspection.
ScalLoc: Continuous Oblique Thermal UAV Geo-Localization with a 2D Satellite Map The paper introduces a framework for continuous oblique thermal drone geo-localization using 2D satellite maps in navigation-denied settings. Preprint	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Thermal-to-satellite geo-localization, not general vision-language or multimodal models.

9,661–9,720 of 17,313← Previous Next →