Gonçalo Paulo

28.8k LoRAs (Olmo-3-7B, Qwen3-8B, Qwen2.5-7B) fine-tuned on incorrect-advice data filtered by attribution scores from many scorer models.