Qwen3-4B-MBT-S/README.md

base_model, language, library_name, license, pipeline_tag, tags

base_model	language	library_name	license	pipeline_tag	tags
Qwen/Qwen3-4B	en	transformers	apache-2.0	text-generation	metacognitive-behavioral-tuning multi-hop-qa reasoning sft grpo MBT-S

Qwen3-4B · MBT-S

MBT-S (main table) — final checkpoint (SFT → GRPO). Base: Qwen/Qwen3-4B. Paper: Metacognitive Behavioral Tuning of Large Language Models for Multi-Hop Question Answering.

• Method: MBT-S (Synthesis): base is SFT'd on gpt-oss-120b-synthesized 5-phase metacognitive traces, then GRPO.
• Base model: Qwen/Qwen3-4B
• Training: SFT (LR 1e-4, BS 128, HotpotQA) → GRPO
• Benchmarks: HotpotQA (ID), MuSiQue / 2WikiMultiHopQA (OOD)

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "metacognitive-behavioral-tuning/Qwen3-4B-MBT-S"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo, dtype="bfloat16", device_map="auto")