This model was produced by first merging Qwen/Qwen3-8B-Base with Qwen/Qwen3-8B, OpenDataArena/Qwen3-8B-ODA-Math-460k, mlabonne/Qwen3-8B-abliterated using the DARE MergeKit method (dare_ties). AIM was then applied to that merged parent using calibration examples from HuggingFaceFW/fineweb-edu (sample-10BT).
