ludwigw/gemma-transitivity-baseline

Gemma Transitivity Baseline (Collapsed)

Standard fine-tuned Gemma 270M-IT on transitivity task without semantic loss. Exhibits catastrophic model collapse — always predicts "Yes" (27.7% accuracy). Provided for reproducibility and as a negative baseline.

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("ludwigw/gemma-transitivity-baseline")
tokenizer = AutoTokenizer.from_pretrained("ludwigw/gemma-transitivity-baseline")

Citation

@article{deshmukh2026semantic,
  title={On Semantic Loss Fine-Tuning Approach for Preventing Model Collapse in Causal Reasoning},
  author={Deshmukh, Pratik and Gupta, Atirek},
  year={2026}
}