Nôm Labby Twohearts Technologies

Nôm Lab — GoiBERT and Vietnamese AI research by Twohearts Technologies.

Sovereign AI / Rooted in Vietnam

01 / Chữ Nôm

Language is
living memory.

Before a word becomes data,
it carries a world.

𠊛ngườiperson · people
Scroll to transform
Follow a word
A note on our name
𡨸
喃

Rooted in Vietnam. Open to the world.

Our future deserves
our own voice.

Chữ Nôm gave written form to Vietnamese stories, knowledge and lived experience. Our name carries that spirit forward.

Nôm Lab is the research lab by Twohearts Technologies. Our ambition is a sovereign AI future: intelligence shaped by Vietnamese language, knowledge and culture, with the capability to build it here.

Our work begins with GoiBERT, a family of compact Vietnamese language encoders—bringing useful intelligence closer to the language people actually live in.

The journey above is an artistic interpretation: one Vietnamese language, two writing systems, and the possibilities of AI. It is not a demonstration of historical-script recognition.

Guided by care, honesty and curiosity. We build to help people understand one another, share knowledge and create lasting value together.

From possibility to practiceResearch note 001 / October 2026

Our latest research

GoiBERT

Research preprint

Intelligence doesn’t
have to be enormous.

GoiBERT explores compact Vietnamese language encoders through two-stage knowledge distillation. The released backbones process raw Vietnamese without an external word-segmentation step, ready for task-specific fine-tuning.

GoiBERT: Compact Vietnamese Language Models via Two-Stage Knowledge Distillation

DownloadRead the paper Explore the code About GoiBERT
GoiBERT-S base in the paper26.4Mparameters
Hugging Face GoiBERT-M large in the paper49.4Mparameters
Hugging Face
Authors & citation ＋

Quang Anh Kiet Tran, Nam Anh Tran, Manh Hung Trinh, Thanh Danh Le, Khang Nhat Hoang Vo, and Huynh Dai Nhan Tran.

GoiBERT: Compact Vietnamese Language Models via Two-Stage Knowledge Distillation. arXiv:2610.30067, 2026.

DOI

Encoder backbones require task-specific fine-tuning. Use is subject to commercial conditions in the model license.

The open notebookFollow an idea to its source

Follow the work.
Further.

01 The paper

02 The models

03 The making

04 The references

The peopleCùng nhau

Curiosity brings us together.
Care guides what we build.

Nôm

This space is for the researchers, builders and collaborators behind Nôm Lab.

Team portraits and stories to come.

Vietnamese roots. A shared future.

The next chapter is a
shared one.

Let’s make something meaningful hello@nomlab.vn
Nôm Labby Twohearts TechnologiesTừ Việt Nam, cùng thế giới.
Nôm Lab / Index
Notes on the artwork

Real characters.
Carefully carried.

Chữ Nôm and quốc ngữ are two ways of writing Vietnamese. Each pairing in this work is an attested reading. Characters can have other readings and meanings in context.

Architectural memories

The floating silhouettes are small artistic studies of places across Vietnam.

The last act presents illustrative associations, not model-generated results. GoiBERT is not presented here as a historical-script recognizer.

Historic glyphs use a subset of Nom Na Tong, from the Vietnamese Nôm Preservation Foundation, under the MIT license.

Read about Chữ Nôm