Cover Image

Beyond Semantics: Disentangling Information Scope in Sparse Autoencoders for CLIP

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Ro, Yusung, Choi, Jaehyun, Kim, Junmo
Format:	Preprint
Published:	2026
Subjects:	Computer Vision and Pattern Recognition
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Cross-Domain Semantic Segmentation on Inconsistent Taxonomy using VLMs
by: Lim, Jeongkee, et al.
Published: (2024)

Learning Neural Deformation Representation for 4D Dynamic Shape Generation
by: Han, Gyojin, et al.
Published: (2026)

Modeling Stereo-Confidence Out of the End-to-End Stereo-Matching Network via Disparity Plane Sweep
by: Lee, Jae Young, et al.
Published: (2024)

Stereo-Matching Knowledge Distilled Monocular Depth Estimation Filtered by Multiple Disparity Consistency
by: Ka, Woonghyun, et al.
Published: (2024)

PRISM: Video Dataset Condensation with Progressive Refinement and Insertion for Sparse Motion
by: Choi, Jaehyun, et al.
Published: (2025)

AH-OCDA: Amplitude-based Curriculum Learning and Hopfield Segmentation Model for Open Compound Domain Adaptation
by: Choi, Jaehyun, et al.
Published: (2024)

Self-supervised One-Stage Learning for RF-based Multi-Person Pose Estimation
by: Shin, Seunghwan, et al.
Published: (2025)

Unlocking the Capabilities of Masked Generative Models for Image Synthesis via Self-Guidance
by: Hur, Jiwan, et al.
Published: (2024)

Frequency-Aware Token Reduction for Efficient Vision Transformer
by: Lee, Dong-Jae, et al.
Published: (2025)

Sparse Autoencoders enable Robust and Interpretable Fine-tuning of CLIP models
by: Morelli, Fabian, et al.
Published: (2026)

Interpreting CLIP with Hierarchical Sparse Autoencoders
by: Zaigrajew, Vladimir, et al.
Published: (2025)

ConceptPrism: Concept Disentanglement in Personalized Diffusion Models via Residual Token Optimization
by: Kim, Minseo, et al.
Published: (2026)

Do Vision Models Encode Object-Level Semantic Relatedness? A Cognitive Psychology-Inspired Benchmark
by: Lee, Hansang, et al.
Published: (2017)

Dream to Generalize: Zero-Shot Model-Based Reinforcement Learning for Unseen Visual Distractions
by: Ha, Jeongsoo, et al.
Published: (2025)

DAM: Domain-Aware Module for Multi-Domain Dataset Condensation
by: Choi, Jaehyun, et al.
Published: (2025)

Self-supervised Transformation Learning for Equivariant Representations
by: Yu, Jaemyung, et al.
Published: (2025)

Self-Predictive Dynamics for Generalization of Vision-based Reinforcement Learning
by: Kim, Kyungsoo, et al.
Published: (2025)

SPACE-CLIP: Spatial Perception via Adaptive CLIP Embeddings for Monocular Depth Estimation
by: Cho, Taewan, et al.
Published: (2026)

Efficiently Disentangling CLIP for Multi-Object Perception
by: Rawlekar, Samyak, et al.
Published: (2025)

Remote Sensing Large Vision-Language Model: Semantic-augmented Multi-level Alignment and Semantic-aware Expert Modeling
by: Park, Sungjune, et al.
Published: (2025)

Concept Steerers: Leveraging K-Sparse Autoencoders for Test-Time Controllable Generations
by: Kim, Dahye, et al.
Published: (2025)

SSG: Scaled Spatial Guidance for Multi-Scale Visual Autoregressive Generation
by: Shin, Youngwoo, et al.
Published: (2026)

Instruct-4DGS: Efficient Dynamic Scene Editing via 4D Gaussian-based Static-Dynamic Separation
by: Kwon, Joohyun, et al.
Published: (2025)

ConceptScope: Characterizing Dataset Bias via Disentangled Visual Concepts
by: Choi, Jinho, et al.
Published: (2025)

ReCLIP++: Learn to Rectify the Bias of CLIP for Unsupervised Semantic Segmentation
by: Wang, Jingyun, et al.
Published: (2024)

B4DL: A Benchmark for 4D LiDAR LLM in Spatio-Temporal Understanding
by: Choi, Changho, et al.
Published: (2025)

Inspecting Explainability of Transformer Models with Additional Statistical Information
by: Nguyen, Hoang C., et al.
Published: (2023)

Cross-Axis Feature Fusion with Joint-Wise Motion Difference Prediction for Text-Based 3D Human Motion Editing
by: Han, Gyojin, et al.
Published: (2026)

Boundary-Recovering Network for Temporal Action Detection
by: Kim, Jihwan, et al.
Published: (2024)

Deciphering the Role of Representation Disentanglement: Investigating Compositional Generalization in CLIP Models
by: Abbasi, Reza, et al.
Published: (2024)

Query-Calibrated Segmental Admission for Descriptor-Agnostic LiDAR Loop Closure in Repetitive Environments
by: Kim, Jaehyun, et al.
Published: (2025)

The Rescue Effect: Spatio-Semantic Early Exit Bypasses Quantization Collapse in CLIP
by: Nam, Kahyeon, et al.
Published: (2026)

Causal Interpretation of Sparse Autoencoder Features in Vision
by: Han, Sangyu, et al.
Published: (2025)

CorrCLIP: Reconstructing Patch Correlations in CLIP for Open-Vocabulary Semantic Segmentation
by: Zhang, Dengke, et al.
Published: (2024)

CLIP-SENet: CLIP-based Semantic Enhancement Network for Vehicle Re-identification
by: Lu, Liping, et al.
Published: (2025)

Disentangled Diffusion Autoencoder for Harmonization of Multi-site Neuroimaging Data
by: Ijishakin, Ayodeji, et al.
Published: (2024)

CLIP-MUSED: CLIP-Guided Multi-Subject Visual Neural Information Semantic Decoding
by: Zhou, Qiongyi, et al.
Published: (2024)

Leveraging Image Augmentation for Object Manipulation: Towards Interpretable Controllability in Object-Centric Learning
by: Kim, Jinwoo, et al.
Published: (2023)

Disentanglement with Factor Quantized Variational Autoencoders
by: Baykal, Gulcin, et al.
Published: (2024)

SRL-CLIP: Efficient CLIP Video Adaptation via Structured Semantic Role Labels
by: Singh, Darshan, et al.
Published: (2024)