latency-sensitive-bench/benchmark-models

Latency-Sensitive Bench models

Inference-ready zero-latency teachers and VLA policies for the supported benchmark tasks.

Layout

• zero-latency/mikasa-intercept-grab-fast/small-policy/ppo-v1: accepted PPO teacher and strict evaluation.
• zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwenoft-v1: StarVLA QwenOFT 5k inference checkpoint, portable config, statistics, and zero-latency evaluation.
• zero-latency/humanoidbench-balance-simple/small-policy/fasttd3-v1: accepted FastTD3 teacher and portable train/eval configs.
• zero-latency/humanoidbench-balance-simple/vla/gr00t-n1.7-10k-v1: official GR00T N1.7 10k inference bundle and zero-latency evaluation.

Extreme Parkour and HAIC artifacts are intentionally not published until their current VLA pipelines pass their documented gates.

Historical source repositories remain unchanged: MIKASA model revision 38327e16f9b33cf5f827b54a606cb4c22d4dfbfc and Balance model revision 734dcedcafbcbffab7f8665d09de42a9c8416d38.

See the repository artifact registry for exact task contracts and reproduction commands.