CLAIAug 7, 2025

H-Net++: Hierarchical Dynamic Chunking for Tokenizer-Free Language Modelling in Morphologically-Rich Languages

arXiv:2508.05628v11 citationsh-index: 9
Originality Incremental advance
AI Analysis

This addresses tokenizer-free language modeling for morphologically-rich languages like Persian, offering an incremental improvement with specialized handling.

The paper tackled the computational challenges of byte-level language models in morphologically-rich languages by proposing H-NET++, a hierarchical dynamic-chunking model, achieving state-of-the-art results including a 0.159 BPB reduction (12% better compression) and a 5.4pp gain on ParsGLUE.

Byte-level language models eliminate fragile tokenizers but face computational challenges in morphologically-rich languages (MRLs), where words span many bytes. We propose H-NET++, a hierarchical dynamic-chunking model that learns linguistically-informed segmentation through end-to-end training. Key innovations include: (1) a lightweight Transformer context-mixer (1.9M parameters) for cross-chunk attention, (2) a two-level latent hyper-prior for document-level consistency, (3) specialized handling of orthographic artifacts (e.g. Persian ZWNJ), and (4) curriculum-based training with staged sequence lengths. On a 1.4B-token Persian corpus, H-NET++ achieves state-of-the-art results: 0.159 BPB reduction versus BPE-based GPT-2-fa (12% better compression), 5.4pp gain on ParsGLUE, 53% improved robustness to ZWNJ corruption, and 73.8% F1 on gold morphological boundaries. Our learned chunks align with Persian morphology without explicit supervision, demonstrating that hierarchical dynamic chunking provides an effective tokenizer-free solution for MRLs while maintaining computational efficiency.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes