NeoteAI/n0-twam-neosim-delta

N0-TWAM post-training — NeoSim 12 tasks (4 single-arm + 8 dual-arm) / pi05_delta (horizon delta)

Multi-task post-trained checkpoint of the N0-TWAM (wan_twam) release tree, step_10000 (final).

Task pool	NeoSim 12 tasks (4 single-arm + 8 dual-arm)
Action space	pi05_delta (horizon delta)
Base	pretrain_mot_umi_mixed/checkpoint_step_16500_r42 (MoT narrow, local-tactile off)
Recipe	LocalTactile flip-on "current", h12 x apf12, MoT narrow experts (action/tactile 1024, ffn 4096), tactile drops 0, lr 1e-4 cosine, 10000 steps
Tactile	GelSight rgb (marker-less)
Norm	per-task q01/q99 (per_robot), NOT a pooled average

Serving note. Multi-task checkpoints must be served with the per-task norm / camera+tactile keys / action channels of the task being evaluated — the pooled envelope in train_meta.json is an unreachable fallback and would de-normalize actions at the wrong scale. Use the ar_server config (AR_SERVE_RUN / AR_SERVE_TASK env); prompts verbatim from the task roster.

Contents: transformer/ (config.json + safetensors, local_tactile tensors = 18) and train_meta.json (training snapshot).