Research map / Label-efficient and robust learning / Recognition and long-tail
Towards Reconfigurable Visual Feature Compression
Jiahang Zhang, Wenhan Yang, Minghao Liu, Jiaying Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use adapters instead of fine-tuning full F_front (ablation) | RFC · PASCAL-Context, 5 requested tasks | Segmentation mIoU: 76.98 -> 77.85 | Improved |
| add recomposition (single forward) (ablation) | RFC · PASCAL-Context, 5 requested tasks | Segmentation mIoU: 77.85 -> 77.43 | Hurt |
| add consolidation coding (ablation) | RFC · PASCAL-Context, 5 requested tasks | bit-rate (bpp): 0.2675 -> 0.1253 | Improved |
| apply RFC over hyperprior FC with split at 2nd layer | RFC · PASCAL-Context | bit-rate (bpp): 0.1566 -> 0.0838 | Improved |
| apply RFC over hyperprior FC with split at 6th layer | RFC · PASCAL-Context | Segmentation mIoU: 75.85 -> 76.26 | Improved |
| apply RFC over hyperprior FC with split at 6th layer | RFC · PASCAL-Context | bit-rate (bpp): 0.1841 -> 0.0771 | Improved |
| use one factorized reconfigurable model instead of multiple single-task models | RFC · PASCAL-Context F-request | bit-rate (bpp): 0.0976 -> 0.0725 | Improved |
| split at 6th layer instead of 2nd layer (ablation) | RFC · PASCAL-Context | bit-rate (bpp): 0.0838 -> 0.0771 | Improved |
Closest papers by meaning
- When MLLMs Meets Compression Distortion: A Coding Paradigm Tailored to MLLMs (ICLR 2026 · similarity 0.85)
- ApET: Approximation-Error Guided Token Compression for Efficient VLMs (CVPR 2026 · similarity 0.85)
- Prune Redundancy, Preserve Essence: Vision Token Compression in VLMs via Synergistic Importance-Diversity (ICLR 2026 · similarity 0.84)
- Knowledge Transfer from Vision Foundation Models for Efficient Training of Small Task-specific Models (ICML 2024 · similarity 0.84)
- VoCo-LLaMA: Towards Vision Compression with Large Language Models (CVPR 2025 · similarity 0.83)
- Variation-aware Vision Token Dropping for Faster Large Vision-Language Models (CVPR 2026 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
