---
base_model: unsloth/Qwen2.5-0.5B-Instruct-bnb-4bit
tags:
- financial-sentiment
- grpo
- rl
- lora
library_name: peft
---

# Financial Sentiment GRPO (baseline)

Fine-tuned with reinforcement learning on Financial PhraseBank (`sentences_50agree`).
AI League #5 Problem Statement 2 POC.

- `loss_type=grpo`, `scale_rewards=True` (standard GRPO baseline)
- Verifiable label reward in `<answer>` tags
- Base model: `unsloth/Qwen2.5-0.5B-Instruct-bnb-4bit`