Brie v2 Qwen 2.5 3B — merged distribution
This is a standalone full-weight distribution produced by merging the
closestfriend/brie-v2-3b LoRA adapter into
Qwen/Qwen2.5-3B-Instruct with PEFT's merge_and_unload() workflow.
It is a deployment convenience for the adapter lineage, not a separately evaluated result. The paper's raw evaluation provenance and canonical model reference point to the adapter repository: https://huggingface.co/closestfriend/brie-v2-3b.
Paper: Human-Curated Data Authoring with LLMs: A Small-Data Approach to Domain Adaptation.
Lineage
Qwen/Qwen2.5-3B-Instruct
+ closestfriend/brie-v2-3b (LoRA adapter)
-> closestfriend/brie-v2-qwen2.5-3b (merged full weights)
The paper's reported Qwen 3B metrics describe outputs generated through the canonical adapter evaluation pipeline. Consult the adapter card and the paper's provenance ledger for the reconciled results and denominator policy; do not count this merged repository as an additional model experiment.
Use
Unlike the adapter repository, this checkpoint loads directly through Transformers and does not require PEFT at inference time.
from transformers import AutoModelForCausalLM, AutoTokenizer
model_id = "closestfriend/brie-v2-qwen2.5-3b"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
This is a research artifact trained for an English-language philosophical and creative register. Verify factual or high-stakes output independently.
Resources
- Canonical adapter: https://huggingface.co/closestfriend/brie-v2-3b
- Paper and versions: https://doi.org/10.5281/zenodo.17657737
- Public method/code mirror: https://github.com/closestfriend/efficient-domain-adaptation
- Base model: https://huggingface.co/Qwen/Qwen2.5-3B-Instruct
Citation
@misc{karman_brie_domain_adaptation,
author = {Karman, Hunter N.},
title = {Human-Curated Data Authoring with LLMs: A Small-Data Approach to Domain Adaptation},
doi = {10.5281/zenodo.17657737},
url = {https://doi.org/10.5281/zenodo.17657737}
}