Research map / Continual learning / Continual fine-tuning of LLMs
Towards Backward-Compatible Continual Learning of Image Compression
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Studies continual learning of neural image compression that extends a pretrained codec to new data or bitrates while keeping backward compatibility with bitstreams from the original model.
Problem
Pretrained neural image compressors cannot be extended to new data or target bitrates without breaking the ability of the original decoder to decode previously encoded bitstreams (backward compatibility).
Approach
Frames image compression extension as continual learning under backward-compatibility constraints, showing end-to-end fine-tuning fails to preserve old decoding ability; the abstract's proposed remedy is not captured in the provided text.
Key results
- Initial findings show that baseline solutions such as end-to-end fine-tuning do not preserve backward compatibility (from abstract).
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fine-tune the compressor end to end on new data or target bitrates | pre-trained neural image compressors · Backward compatibility | backward compatibility | Hurt |
Limitations
- Full text is largely garbled; method details and experiments could not be verified.
Experimental setup
Baselines: end-to-end fine-tuning
Closest papers by meaning
- Online Learned Continual Compression with Adaptive Quantization Modules (ICML 2020 · similarity 0.89)
- GLIC: General Format Learned Image Compression (AAAI 2025 · similarity 0.86)
- Tensor Decomposition Based Memory-Efficient Incremental Learning (ICML 2025 · similarity 0.86)
- Salient Frequency-aware Exemplar Compression for Resource-constrained Online Continual Learning (AAAI 2025 · similarity 0.86)
- Controllable Distortion-Perception Tradeoff Through Latent Diffusion for Neural Image Compression (AAAI 2025 · similarity 0.85)
- Test-Time Fine-Tuning of Image Compression Models for Multi-Task Adaptability (CVPR 2025 · similarity 0.85)
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.