| language: en | |
| license: mit | |
| tags: | |
| - scibert | |
| - classification | |
| - technical-papers | |
| - machine-learning | |
| # AEGIS SciBERT Technical Classifier | |
| ## Model Description | |
| Fine-tuned SciBERT model for classifying technical papers into research categories. | |
| ## Training Details | |
| - **Base Model**: allenai/scibert_scivocab_uncased | |
| - **Training Samples**: 500 | |
| - **Number of Classes**: 6 | |
| - **Classes**: cs.AI, cs.LG, quant-ph, cs.NE, stat.ML, cs.CV | |
| - **Validation Accuracy**: 1.0000 | |
| ## Usage | |
| ```python | |
| from transformers import AutoModelForSequenceClassification, AutoTokenizer | |
| model = AutoModelForSequenceClassification.from_pretrained("gsstec/aegis-scibert-technical") | |
| tokenizer = AutoTokenizer.from_pretrained("gsstec/aegis-scibert-technical") | |
| # Example inference | |
| text = "Quantum computing algorithms for machine learning" | |
| inputs = tokenizer(text, return_tensors="pt") | |
| outputs = model(**inputs) | |
| predictions = torch.softmax(outputs.logits, dim=-1) | |
| ``` | |
| ## Classes | |
| - `cs.AI`: Class 0 | |
| - `cs.LG`: Class 1 | |
| - `quant-ph`: Class 2 | |
| - `cs.NE`: Class 3 | |
| - `stat.ML`: Class 4 | |
| - `cs.CV`: Class 5 |