Moonlight556/sokrates-qwen3-8b-prontoqa-sft-optionized

SOKRATES: Qwen3-8B PrOntoQA SFT (Optionized)

Supervised fine-tuned Qwen3-8B for logical reasoning using the optionized Thought/Action format.

Performance

Metric	Value
PrOntoQA Accuracy	93.3%

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "Moonlight556/sokrates-qwen3-8b-prontoqa-sft-optionized",
    torch_dtype="bfloat16",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(
    "Moonlight556/sokrates-qwen3-8b-prontoqa-sft-optionized"
)

Related Models

• DPO Iter 1: sokrates-qwen3-8b-prontoqa-oak-dpo-iter1 (96.8%)
• DPO Iter 2: sokrates-qwen3-8b-prontoqa-oak-dpo-iter2 (98.1%)
• DPO Iter 3: sokrates-qwen3-8b-prontoqa-oak-dpo-iter3 (98.2% - best)