---
language: en
license: apache-2.0
base_model: mistralai/Mistral-7B-Instruct-v0.3
tags:
- text-classification
- political-text
- mistral
- lora
pipeline_tag: text-classification
---

# your-username/your-repo-name

**Auto-generated model card** based on the training script.

- base model: mistralai/Mistral-7B-Instruct-v0.3
- labels: DIFFUSE_SUPPORT, SPECIFIC_SUPPORT, PRINCIPLED_OPPOSITION, CLASSICAL_OPPOSITION, NEUTRAL
- max_seq_len: 256
- pooling: last-token pooling (uses attention mask)
- classifier: Linear -> ReLU -> Dropout(0.1) -> Linear
- two-stage fine-tuning: head-only (epochs=2) then LoRA + unfreeze last N (epochs=3)