English

Large Language Model as Token Compressor and Decompressor

Computation and Language 2026-05-14 v2

Abstract

In this paper, we study whether an off-the-shelf LLM can be adapted into a discrete, variable-length token compressor and decompressor for long-context processing. To this end, we design a self-expressive autoencoding framework that fine-tunes a pretrained LLM with lightweight LoRA adapters to map long texts into compact sequences of learned latent codes, termed Z-tokens, and to decode them back into natural language or task outputs. The resulting representation is content-adaptive: less predictable or information-dense segments can receive more Z-tokens, while redundant regions can be represented more compactly through a budget-aware length regularizer. Our method is evaluated on long-context datasets such as Wikipedia, CNN/DailyMail, HotpotQA, and QuALITY, showing that it preserves reconstruction quality and downstream performance while reducing effective context length, generation-stage memory usage, and end-to-end latency. This simple design supports both direct decoding from compressed contexts and autoregressive generation in the Z-token space, providing a practical interface for efficient long-context inference.

Keywords

Cite

@article{arxiv.2603.25340,
  title  = {Large Language Model as Token Compressor and Decompressor},
  author = {Wenbing Li and Yiran Wang and Zikai Song and Jielei Zhang and Tianhao Zhao and Junkai Lin and Wei Yang},
  journal= {arXiv preprint arXiv:2603.25340},
  year   = {2026}
}
R2 v1 2026-07-01T11:39:06.089Z