# Cartesia — Hierarchical modeling

- Company: Cartesia (cartesia.ai)
- Announced: 2025-07-11
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.cartesia.ai/blog/hierarchical-modeling
- Record: https://forck.live/items/8016-hierarchical-modeling
- Subject: Sonic / Ink
- Models affected: H-Net, H-Net 2-stage XL, H-Net 1-stage XL, H-Net 1-stage L

Cartesia announced a research collaboration on hierarchical networks (H-Nets), a new architecture that learns to segment and compress raw data into meaningful concepts. H-Nets scale better with data than Transformers with BPE tokenization, are more robust to input perturbations, and can be stacked for deeper hierarchies. The company released checkpoints for H-Net models on HuggingFace.

## Evidence

Verbatim from https://www.cartesia.ai/blog/hierarchical-modeling:

> We’re excited to announce our latest research collaboration on hierarchical networks (H-Nets) , a new architecture that natively models hierarchy from raw data.

---

Record: https://forck.live/items/8016-hierarchical-modeling
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md