AWS on Cosmos 3: full Nano/Super checkpoints tested in BF16; H100 8-GPU SFT tested by NVIDIA. The post provides no benchmark numbers.
F4R diagnoses real failures with a VLM agent, rebuilds them in sim and trains around them; authors report 90.0% OOD success, 18.75 pts over Targeted BC.
ActionUNet adds a temporal U-Net and SIREN decoder to a frozen VLA; authors report π0.5 gains of 9.8/6.1/11.4 points in robustness tests, 6.46% slower calls.
Preprint: 4-bit formats on Orin NX differ 36.8× in step energy; IQ4_NL has the lowest energy and resident memory, while NF4 and INT4 also run on-board.
Match a raw answer to a SmolVLA trace. Explanation: not_assessed; unassessed facts give exit 1. Keep format failures unchanged.
Recompute a model-upgrade verdict from pinned, verified records: the total rose, yet 10 blocking checks trace to format or schema failures in two cases.
Frozen VLA plus a sim-only residual on VLM latents aligned by an optimal-transport mapper; authors report real-robot gains in all 8 task–backbone pairs.
π0.5-derived waypoint VLA: plans took 57 serial passes, endpoint erasure didn't hurt success. Authors cut robot planning 1094→125 ms; NGM makes plans count.
Sequential robot policy comparison via safe anytime-valid inference, for non-binary metrics too; authors report up to 70% less evaluation vs batch tests.
14 items from the Physical AI Radar's September archive, numbers checked: VLA evaluation starts testing recovery; latency becomes a formally reported metric.
7 primary sources: recovery benchmarks, sequential testing, post-quantization closed-loop checks, sim fidelity. 98.83% initial success, only 48% recovery.
Half-day SFT, RL post-training, learning from deployment: 7 radar sources outline VLA post-training, and the risk that online RL silently breaks single tasks.
Sign in to chat with the assistant.
Sign in