Legal-domain fine-tune
According to the model card, the model is fine-tuned from microsoft/harrier-oss-v1-0.6b for the Vietnamese legal domain.
Open Source Model Profile · mainguyen9
vietlegal-harrier-0.6b
vietlegal-harrier-0.6b is a 0.60B-parameter Vietnamese legal-domain embedding model from mainguyen9. According to the model card, it is fine-tuned from microsoft/harrier-oss-v1-0.6b with 1024-dim output and cosine similarity.
- Publisher
- mainguyen9
- Task
- sentence-similarity
- Model type
- qwen3
- License
- apache-2.0
- Library
- sentence-transformers
- Publication status
- Approved for indexing
Model overview
vietlegal-harrier-0.6b is published by mainguyen9 as a sentence-similarity model for Vietnamese legal retrieval. Safetensors metadata reports 596,049,920 parameters, and the captured configuration identifies Qwen3Model. According to the model card, the base model is microsoft/harrier-oss-v1-0.6b with maximum sequence length 512 tokens.
Recorded capabilities
1024-dim normalized embeddings
According to the model card, output dimensionality is 1024 with cosine similarity and a normalize step after last-token pooling.
Reported retrieval scores
According to the model card, the model achieves NDCG@10 0.7813, MRR@10 0.7303, and Recall@10 0.9321 on the Zalo AI Legal Text Retrieval benchmark.
Documented comparison table
According to the model card, the reported table places it above vietlegal-e5 at 0.7310, harrier-oss-v1-0.6b at 0.7210, and multilingual-e5-large at 0.6660 NDCG@10.
Use cases in the source record
- Vietnamese legal document retrieval using sentence embeddings with cosine similarity.
- Legal query-passage matching experiments evaluated on the MTEB ZacLegalTextRetrieval set the card describes.
Limitations and unknowns
- Reported benchmark figures and baseline comparisons are publisher claims from the model card and were not independently verified.
- Training corpus, pipeline, and hardware details are publisher-reported and were not independently verified.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: mainguyen9/vietlegal-harrier-0.6b
Captured: Unknown. Processed: 2026-09-07T19:34:50.995392+00:00.
VietLegal-Harrier-0.6B A Vietnamese legal domain embedding model fine-tuned from microsoft/harrier-oss-v1-0.6b (600M params, 1024-dim). Achieves NDCG@10 = 0.7813 on the Zalo AI Legal Text Retrieval benchmark, outperforming all baselines including our previous vietlegal-e5 model. Benchmark Results Evaluated on MTEB ZacLegalTextRetrieval (61.4K corpus documents, 818 test queries). Model Params Dim NDCG@10 MRR@10 Recall@10 mainguyen9/vietlegal-harrier-0.6b 600M 1024 0.7813 0.7303 0.9321 mainguyen9/vietlegal-e5 (mE5-large) 560M 1024 0.7310 0.6770 0.8972 mainguyen9/vietlegal-harrier-270m 270M 1024 0.7174 0.6636 0.8864 microsoft/harrier-o…
F001F002F003F004F005F006F007F008F009F010F011F012F013F014F015F016F017