[2603.03583] ByteFlow: Language Modeling through Adaptive Byte Compression without a Tokenizer

Computer Science > Computation and Language arXiv:2603.03583 (cs) [Submitted on 3 Mar 2026] Title:ByteFlow: Language Modeling through Adaptive Byte Compression without a Tokenizer Authors:Chunyuan Deng, Sanket Lokegaonkar, Colin Lockard, Besnik Fetahu, Nasser Zalmout, Xian Li View a PDF of the paper titled ByteFlow: Language Modeling through Adaptive Byte Compression without a Tokenizer, by Chunyuan Deng and 5 other authors View PDF HTML (experimental) Abstract:Modern language models still rely on fixed, pre-defined subword tokenizations. Once a tokenizer is trained, the LM can only operate at this fixed level of granularity, which often leads to brittle and counterintuitive behaviors even in otherwise strong reasoning models. We introduce \textbf{ByteFlow Net}, a new hierarchical architecture that removes tokenizers entirely and instead enables models to learn their own segmentation of raw byte streams into semantically meaningful units. ByteFlow Net performs compression-driven segmentation based on the coding rate of latent representations, yielding adaptive boundaries \emph{while preserving a static computation graph via Top-$K$ selection}. Unlike prior self-tokenizing methods that depend on brittle heuristics with human-designed inductive biases, ByteFlow Net adapts its internal representation granularity to the input itself. Experiments demonstrate that this compression-based chunking strategy yields substantial performance gains, with ByteFlow Net outperforming both ...