Abstract:Modern language models are trained almost exclusively on token sequences produced by a fixed tokenizer, an external lossless compressor often over UTF-8 byte sequences, thereby coupling the model to that compressor. This work introduces proxy compression, an alternative training scheme that preserves the efficiency benefits of compressed inputs while providing an end-to-end, raw-byte interface at inference time. During training, a single language model is jointly trained on raw byte sequences and compressed views generated by external compressors; through the process, the model learns to internally align compressed sequences and raw bytes. This alignment enables strong transfer between the two formats, even when training predominantly on compressed inputs that are discarded at inference. Extensive experiments on code language modeling demonstrate that proxy compression substantially improves training efficiency and significantly outperforms pure byte-level baselines given fixed compute budgets. As model scale increases, these gains become more pronounced, and proxy-trained models eventually match or surpass tokenizer approaches, all while operating solely on raw bytes and retaining the inherent robustness of byte-level modeling. Our code is available at this https URL.
| Comments: | ICML 2026 |
| Subjects: | Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2602.04289 [cs.CL] |
| (or arXiv:2602.04289v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2602.04289 arXiv-issued DOI via DataCite |
Submission history
From: Lin Zheng [view email]
[v1]
Wed, 4 Feb 2026 07:36:46 UTC (260 KB)
[v2]
Thu, 14 May 2026 17:33:56 UTC (264 KB)