ELLIS Unit Trento at CVPR 2026: 4 Findings, 16 Posters, and 5 Workshops Presented

Trento, June 2026 — ELLIS Unit Trento has marked a significant milestone at CVPR 2026, showcasing an impressive lineup of 4 findings, 16 poster presentations, and 5 organized workshops.

CVPR is widely recognized as the premier international conference in computer vision and pattern recognition. The remarkable volume of accepted works and leadership roles at this year’s event underscores the international visibility, scientific excellence, and continuous growth of ELLIS Unit Trento’s research activities.

Authored and led by researchers affiliated with the Unit, the presented works spanned critical areas of computer vision, machine learning, and artificial intelligence, reflecting both groundbreaking methodological innovation and real-world impact.

This outstanding contribution highlights the strength of ELLIS Unit Trento’s research ecosystem, its commitment to pushing the boundaries of the state of the art, and its vital role within the global AI community.

ELLIS Unit Trento warmly congratulates all the authors and organizers involved for their remarkable efforts and significant scientific contributions.

Below is the complete list of our CVPR 2026 papers and workshops.

Workshops

• • 9th Multimodal Learning and Applications Workshop

  • Pixel-level Video Understanding in the Wild Challenge

  • Workshop on Any-to-any Multimodal Learning 

  • 11th New Trends in Image Restoration and Enhancement Workshop and Challenges 

  • 9th International Workshop on Visual Odometry and Computer Vision Applications Based on Location Clues 

Findings

• Finetune Like You Pretrain: Boosting Zero-shot Adversarial Robustness in Vision-language Models

• Large Multimodal Models as General In-Context Classifiers

• Organizing Unstructured Image Collections using Natural Language

• SEM: Sparse Embedding Modulation for Post-Hoc Debiasing of Vision-Language Models

Posters

• • From Weights to Concepts: Data-Free Interpretability of CLIP via Singular Vector Decomposition

  • The Devil Is in Gradient Entanglement: Energy-Aware Gradient Coordinator for Robust Generalized Category Discovery

  • CLIPoint3D: Language-Grounded Few-Shot Unsupervised 3D Point Cloud Domain Adaptation

  • Token Reduction via Local and Global Contexts Optimization for Efficient Video Large Language Models

  • Open-Vocabulary Domain Generalization in Urban-Scene Segmentation

  • PoInit-of-View: Poisoning Initialization of Views Transfers Across Multiple 3D Reconstruction Systems

  • TerraScope: Pixel-Grounded Visual Reasoning for Earth Observation

  • Chorus: Multi-Teacher Pretraining for Holistic 3D Gaussian Scene Encoding

  • Generalizable Knowledge Distillation from Vision Foundation Models for Semantic Segmentation

  • Deconstructing the Failure of Ideal Noise Correction: A Three-Pillar Diagnosis

  • How to Take a Memorable Picture? Empowering Users with Actionable Feedback

  • Hyperbolic Busemann Neural Networks

  • Specificity-aware reinforcement learning for fine-grained open-world classification

  • StructXLIP: Enhancing Vision-language Models with Multimodal Structural Cues

  • Obstruction reasoning for robotic grasping

  • Efficient Encoder-Free Fourier-based 3D Large Multimodal Model