用视觉压缩长文本效果不佳，直接方法反而更优。
Optical Context Compression Is Just (Bad) Autoencoding
- 将文本嵌入转为图像再压缩，反而损失信息
- 不同压缩比下，视觉方法均不如直接处理
- 适合关注高效文本压缩的开发者
DeepSeek-OCR 表明，少量视觉标记即可重建渲染文本，引发对视觉作为长文本压缩媒介的兴趣。但该方法需将标记嵌入转为像素并压缩，丢弃学习到的表示，再由视觉编码器恢复。我们对比 DeepSeek-OCR 的视觉编码器、近零参数均值池化和学习型分层编码器发现：其并未带来优势。在重建任务中，直接方法在所有压缩比下均匹配或超越视觉方法；在语言建模中，视觉方法仅与截断基线相当，且在各压缩比下均落后于分层编码器。尽管所有压缩方法均优于截断法的事实记忆表现，但视觉方法从未超越最佳直接基线。对光学上下文压缩的热情超出了实证支持。代码与检查点见 https://github.com/ivnle/bad-autoencoding。
原文摘要 · Abstract (English)
DeepSeek-OCR shows that rendered text can be reconstructed from a small number of vision tokens, sparking excitement about using vision as a compression medium for long textual contexts. But this pipeline requires rendering token embeddings to pixels and compressing from there -- discarding learned representations in favor of an image the vision encoder must then recover from. We ask whether this detour helps. Comparing DeepSeek-OCR's vision encoder against near-zero-parameter mean pooling and a learned hierarchical encoder, we find it does not. For reconstruction, simple direct methods match or surpass vision at every compression ratio. For language modeling, vision performs comparably to truncation -- a baseline that simply discards context -- and loses to the hierarchical encoder at every compression ratio. As expected, all compression methods outperform truncation for factual recall, but vision never surpasses the best direct baseline. The excitement around optical context compression outpaces the evidence. Code and checkpoints are available at https://github.com/ivnle/bad-autoencoding.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。