{ "base_model": "Qwen/Qwen3.5-4B", "families": [ "Reasoning base" ], "training_rows": 1282, "rows_trained": 5760, "validation_ce": 0.2812701124697924, "minutes": 86.9, "args": { "adapter": "reasoning", "epochs": 3.0, "lr": 0.0001, "rank": 64, "alpha": 128, "grad_accum": 8, "beta": 0.1, "anchor_ratio": 0.5, "limit": 0, "max_steps": 0, "seed": 0, "ce_on": "all", "exclude_variants": "" } }