Abstract
In this work, we aim to discretize the high-dimensional visual representations to bridge the gap with language models - a non-trivial challenge, as existing quantization methods suffer from codebook collapse, failing to scale while preserving semantic coherence. We identify the root cause as metric mismatch: standard Euclidean codebook objectives are fundamentally misaligned with the anisotropic geometry of representation space, leading to codebook embeddings with high-variance magnitude scales and uneven angular distributions that hinder scalability. To address this, we propose Hyper-Spherical Quantization (HSQ), which decouples semantic content from feature magnitude via angular routing, preventing code assignment from being dominated by scale rather than meaning. The resulting discrete Representation Autoencoder (dRAE) achieves high-fidelity reconstruction while preserving semantic integrity and supporting scalable codebook budget. Extensive experiments demonstrate consistent performance gains as the vocabulary size scales to 131{,}072, along with 100\% codebook utilization, simplified training pipeline, and strong performance across understanding and generation tasks.
Community
We propose dRAE, a visual tokenizer built on Hyper-Spherical Quantization (HSQ). By assigning codes via cosine similarity and updating embeddings in tangent space, HSQ aligns the quantization objective with the intrinsic geometry of pre-trained vision encoder features (e.g., SigLIP2, DINOv2), achieving exceptionally high codebook utilization and strong semantic alignment.
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- VideoRAE: Taming Video Foundation Models for Generative Modeling via Representation Autoencoders (2026)
- HiTokSR: A Coarse-to-Fine Tokenizer with Hierarchical Codebooks for High-Fidelity Real-World Image Super-Resolution (2026)
- MIMFlow: Integrating Masked Image Modeling with Normalizing Flows for End-to-End Image Generation (2026)
- Self-Guidance: Enhancing Neural Codecs via Decoder Manifold Alignment (2026)
- Unified Driving Tokens: Representation- and Geometry-Guided Discrete Tokenizer for Driving World Models and Planning (2026)
- NSVQ: Mitigating Codebook Collapse by Stabilizing Encoder Drift in Vector Quantization (2026)
- GEAR: Guided End-to-End AutoRegression for Image Synthesis (2026)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend
Get this paper in your agent:
hf papers read 2607.22148 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper