TERRA — post-trained policies (LIBERO, and one real-robot arm)

OpenVLA-7B with TERRA latent-action phase 2 (latp2_inv labels), post-trained on LIBERO (flow head, 4 query tokens). One directory per policy: merged weights (model-*.safetensors), action/latent heads (action_decoder.pt), dataset_statistics.json (action un-normalisation), ctrl_meta.json, processor / tokenizer. Evaluate the LIBERO ones with acdc_la.downstream.vla.run_libero_eval_cont --pretrained_checkpoint <dir> (repo Latent_Free, branch newton). A real-robot directory carries its own README with its interface.

dir	suite	demos	standard LIBERO (500 ep.)
libero_10_full	libero_10_no_noops	full	89.0
libero_goal_full	libero_goal_no_noops	full	94.4
piper_pick_can_10k	piper_pick_can (real robot, AgileX Piper)	142	n/a — real robot, see its README