llama-3.2-3b-instruct-unsloth-bnb-4bit-gnd-subjects

This model is an attempt at the shared task LLMs4Subjects (https://sites.google.com/view/llms4subjects-germeval/home) for Subtask 2, to classify subject domains of German and English documents. The task is a multi-class multi-label subject indexing task, with the GND subject labels (over 200,000 in the set). Ihe model is fine-tuned on a dataset of prompt-embedded texts with GND subject classes, derived from the TIBKAT dataset (more about the TIBKAT dataset TIB Open Data Services: https://www.tib.eu/en/services/open-data). The model was used for prompting as a part of the pipeline system for subject indexing and had received the 3rd place for the subtask, with a recall score of 0.2 (K@20) on the test set evaluated by the organisers on CodaBench (https://www.codabench.org/competitions/8373/#/results-tab).

Training and evaluation data

• Training set: https://huggingface.co/datasets/ubffm/linsearch_train_data
• Development set: https://huggingface.co/datasets/ubffm/linsearch_dev_data
• TIBKAT documents with title, abstract and GND subject classes embedded in prompts
• Size: 131k examples (116k training set, 18.7k development set)

How to use

with Huggingface pipeline

from transformers import pipeline

model_id = "ubffm/llama-3.2-3b-instruct-unsloth-bnb-4bit-gnd-subjects"
pipe = transformers.pipeline(
    "text-generation",
    model=model_id,
    tokenizer=model_id,
    model_kwargs={"torch_dtype": torch.bfloat16},
    device_map="auto",
)

prompt = "Your prompt here."
output = pipe(prompt, max_new_tokens=200, do_sample=True, temperature=0.7)
print(output[0]["generated_text"])

Contact and Citation

@inproceedings{ho-2025-ubffm,
    title = "{UBFFM} at the {G}erm{E}val-2025 {LLM}s4{S}ubjects Task: What if we take ``You are an expert in subject indexing'' seriously?",
    author = "Ho, Clara Wan Ching",
    editor = "Wartena, Christian  and
      Heid, Ulrich",
    booktitle = "Proceedings of the 21st Conference on Natural Language Processing (KONVENS 2025): Workshops",
    month = sep,
    year = "2025",
    address = "Hannover, Germany",
    publisher = "HsH Applied Academics",
    url = "https://aclanthology.org/2025.konvens-2.44/",
    pages = "471--478"
}

@misc{ubffm/llama-3.2-3b-instruct-unsloth-bnb-4bit-gnd-subjects,
  title={llama-3.2-3b-instruct-unsloth-bnb-4bit-gnd-subjects},
  author={UBFFM},
  year={2025},
  howpublished={\url{https://huggingface.co/ubffm/llama-3.2-3b-instruct-unsloth-bnb-4bit-gnd-subjects}},
}

Contact email: info@linguistik.de

Uploaded model

• Developed by: c-ho
• License: apache-2.0
• Finetuned from model : unsloth/llama-3.2-3b-instruct-unsloth-bnb-4bit

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.