decider-2b v10 + merged LoRA (rank 64, attention and MLP) with replay trained toward v10's distribution; temperature 1.145, temperature_by_type choice 1.164 / noul 1.624 / score 1.124 (decider-ai 1.4.0). v10 is under the tag v10, v8 under v8.
Card: field types for decider-ai 1.3.0 (confidence, x_p_max)