metadata
base_model: unsloth/Qwen3-1.7B-unsloth-bnb-4bit
license: mit
pipeline_tag: text-generation
tags:
- base_model:adapter:unsloth/Qwen3-1.7B-unsloth-bnb-4bit
- lora
- sft
- transformers
- trl
- unsloth
- citation-preprocessing
- text-refinement
- academic-text-processing
inference:
parameters:
temperature: 0.6
top_p: 0.95
max_new_tokens: 512
ollama:
parameters:
num_ctx: 1024
temperature: 0.6
top_p: 0.95
SemanticCite-Refiner-Qwen3-1B
A fine-tuned Qwen3-1.7B model specialized for preprocessing citation text. This model removes reference markers, author names, and publication identifiers while converting author-centered statements to fact-centered statements for improved citation verification.
Model Details
Model Description
This model is designed to preprocess citation text by cleaning and standardizing it for downstream verification tasks. It removes reference markers (e.g., [1], Smith 2020, et al.), converts author-centered statements to fact-centered statements using passive voice, while maintaining all numerical values and factual details.
- Developed by: Sebastian Haan
- Model type: Causal Language Model (Fine-tuned)
- Language(s) (NLP): English
- License: MIT
- Finetuned from model: unsloth/Qwen3-1.7B-unsloth-bnb-4bit
Uses
Direct Use
This model is intended for:
- Preprocessing citation text for academic verification systems
- Cleaning and standardizing citation statements
- Converting author-centric to fact-centric statements
- First stage in citation verification pipelines
Out-of-Scope Use
This model should not be used for:
- General text summarization or rewriting
- Legal document processing
- Medical text processing
- Creative writing or content generation