---
base_model: Qwen/Qwen3-0.6B
language:
- en
library_name: transformers
license: apache-2.0
pipeline_tag: text-generation
tags:
- metacognitive-behavioral-tuning
- multi-hop-qa
- reasoning
- sft
- grpo
- gpt-oss-distill
---

# Qwen3-0.6B · gpt-oss-distill

gpt-oss-distill (appendix) — **final checkpoint (SFT → GRPO)**. Base: [`Qwen/Qwen3-0.6B`](https://huggingface.co/Qwen/Qwen3-0.6B).
Paper: *Metacognitive Behavioral Tuning of Large Language Models for Multi-Hop Question Answering*.

- **Method**: gpt-oss-distill: naive distillation of teacher (gpt-oss-120b) raw traces via SFT, then GRPO.
- **Base model**: `Qwen/Qwen3-0.6B`
- **Training**: SFT (LR 1e-4, BS 128, HotpotQA) → GRPO
- **Benchmarks**: HotpotQA (ID), MuSiQue / 2WikiMultiHopQA (OOD)

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "metacognitive-behavioral-tuning/Qwen3-0.6B-gpt-oss-distill"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo, dtype="bfloat16", device_map="auto")
```