GOIBERT-S AND GOIBERT-M
THIRD-PARTY ATTRIBUTIONS AND NOTICES

Copyright (c) 2026 CÔNG TY TNHH TWOHEARTS TECHNOLOGIES, for its original
GoiBERT contributions. All third-party rights remain with their owners.

Licensor: CÔNG TY TNHH TWOHEARTS TECHNOLOGIES
Vietnam tax code: 0319350288
Commercial licensing contact: nhan@twohearts.vn

Model repositories:
https://huggingface.co/TwoheartsTechnologies/GoiBERT-S
https://huggingface.co/TwoheartsTechnologies/GoiBERT-M

1. BamiBERT

GoiBERT-S and GoiBERT-M were developed through general distillation from
BamiBERT. Student word embeddings were initialized by PCA projection of
BamiBERT word embeddings. The models use the BamiBERT tokenizer.

Upstream repository:
https://huggingface.co/Qualcomm-AI-Research/BamiBERT

Upstream revision:
57bc1340debbe4e348ec549047a763caebe4a977

BamiBERT is released under the BSD-3-Clause-Clear license and the Qualcomm
Responsible AI License. Applicable upstream license terms, copyright
notices and disclaimers remain in effect. The GoiBERT Community Model
License applies to Twohearts' licensable contributions and does not replace
or waive rights or obligations under these third-party licenses.

Applicable upstream license texts:
licenses/BSD-3-Clause-Clear.txt
licenses/Qualcomm-RAIL.txt

BSD-3-Clause-Clear reference:
https://spdx.org/licenses/BSD-3-Clause-Clear.html

Qualcomm Responsible AI License:
https://www.qualcomm.com/site/responsible-ai-license

BamiBERT reference:
Dat Quoc Nguyen, Thinh Pham, Chi Tran, and Linh The Nguyen.
BamiBERT: A New BERT-based Language Model for Vietnamese.
arXiv:2607.02259, 2026.
https://arxiv.org/abs/2607.02259

2. Training-data attribution

The general-distillation corpus includes HPLT 3.0 Vietnamese web text,
Vietnamese Wikipedia text, user-generated text extracted from HPLT, and
unlabeled training-split text from UIT-VSFC, UIT-ViSFD, UIT-ABSA Restaurant,
UIT-ABSA Hotel, ViSpamReviews, ViNLI and PhoNER_COVID19.

HPLT 3.0:
https://huggingface.co/datasets/HPLT/HPLT3.0

UIT-ViSFD:
Luong Luc Phan et al. SA2SL: From Aspect-Based Sentiment Analysis to Social
Listening System for Business Intelligence. KSEM 2021, pages 647-658.
https://doi.org/10.1007/978-3-030-82147-0_53
https://github.com/LuongPhan/UIT-ViSFD

Dataset rights remain with their respective rights holders. This notice
provides attribution and does not grant rights to reproduce or redistribute
the underlying datasets. The model license does not replace applicable
third-party permissions or terms.

3. Trademarks and endorsement

All third-party names and marks belong to their respective owners.
References to BamiBERT, Qualcomm, dataset authors or institutions identify
provenance and do not imply sponsorship, affiliation or endorsement.