1.2 KiB
1.2 KiB
license, language, base_model, pipeline_tag, tags
| license | language | base_model | pipeline_tag | tags | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| mit |
|
|
text-generation |
|
GPT-2 Fine-Tuned for Academic Style Writing
This model is a fine-tuned version of GPT-2 (124M), trained on 500,000 academic abstracts from the arXiv dataset. It was developed as part of the ECS 271 final project at UC Davis.
Intended Use
Designed to generate text with academic tone and structure for use in educational and research-oriented applications such as writing assistants or tutors.
Training Details
- Base model:
gpt2 - Method: Full-parameter fine-tuning
- Dataset: 500K arXiv abstracts
- Epochs: 1.2
- GPU: RTX 5070 Ti (~2 hours)
Limitations
- May hallucinate facts
- Not citation-aware
- English-only
- No factual grounding
Example
Prompt:
"Transformer models have impacted the field of education by..."
Output:
"Transformer models have impacted the field of education by enabling new forms of intelligent tutoring systems and automatic curriculum generation. Their ability to understand context has allowed for more personalized and scalable academic tools."
More Info
Main Project: Github