| --- |
| library_name: peft |
| tags: |
| - dementor |
| - imitation-disguise |
| - lora |
| - chatbot_arena |
| base_model: adamo1139/aya-expanse-8b-ungated |
| --- |
| |
| # dpo_chatbot_arena_aya-expanse-8b_as_gemma-4-e4b_seed42 |
|
|
| CURRENT Dementor imitation (disguise) LoRA adapter — dataset **chatbot_arena**, **seed 42**. |
| |
| - **Method:** DPO |
| - **Source model (fine-tuned / disguised):** `aya-expanse-8b` (base: `adamo1139/aya-expanse-8b-ungated`) |
| - **Target model being imitated:** `gemma-4-e4b` |
| - **Dataset:** chatbot_arena (benign) | **Seed:** 42 |
|
|
| This adapter trains the source model to imitate the target model's style on the benign |
| chatbot_arena corpus. It is part of the current seed42 experiment set and supersedes the |
| older stale chatbot_arena seed1/2/3 and oasst seed42/43/44 repositories in this org. |
| Registry key == repo id == `dpo_chatbot_arena_aya-expanse-8b_as_gemma-4-e4b_seed42`. |
|
|
