altaidevorg/gemma-judge-v0.1

Gemma Judge

This model is a compact yet highly capable LLM-as-a-judge model, fine-tuned from Gemma3 +B. It can be used for both direct feedback evaluations and A/B preference evaluations.

It is obtained by merging two models separately fine-tuned on feedback and preference tasks.

Models Merged

The following models were included in the merge:

• altaidevorg/gemma-judge-feedback-v0.1
• altaidevorg/gemma-judge-preferences-v0.1

Benchmarks

Model	Benchmark	Exact Match Accuracy (%)	Pearson r	Spearman ρ	Accuracy (%) (Pairwise)	Notes
🟪 altaidevorg/gemma-judge-v0.1	Feedback Bench / Preference Bench	73.0 / –	0.9198 / –	0.9210 / –	94.54	Strong unified performance across both tasks
🟨 Prometheus 2 (8×7B) (Kim et al., 2024)	Feedback Bench / Preference Bench	– / –	≈ 0.898 / –	≈ 0.90 / –	90.65	–

---

🧾 License

This model is released under the Apache 2.0 License.
However, because it is derived from Google’s Gemma 3, your use of this model must also comply with the Gemma Terms of Use.

By using this model, you agree to:

• Follow Google’s Gemma Model Terms of Use, including restrictions on misuse and redistribution.
• Attribute Google as the original provider of the Gemma 3 base model.

For full details, see: https://ai.google.dev/gemma/terms

---