Align Once to Explain: Feature Alignment for Scalable B-cosification of Foundational Vision Transformers
cs.CV
Submitted: 2026-11-02
Updated: 2026-11-02
Code: https://github.com/shrebox/B-cosification
Terminology
Sources
- Democratizing Contrastive Language-Image Pre-training: A CLIP Benchmark of Data, Model, and Supervision
- Distilling the Knowledge in a Neural Network
- Captum: A unified and generic model interpretability library for PyTorch
- Fine-Grained Visual Classification of Aircraft
- DINOv3
- Gemma 2: Improving Open Language Models at a Practical Size
- SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features
- Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer
- Rethinking Centered Kernel Alignment in Knowledge Distillation
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs