Does OpenVLA-OFT run on I2RT YAM?

Capabilities

Caveats

• Community-reported only; not documented upstream or maintainer-tested (evidence is a baseline in the MolmoAct2 paper, not the OpenVLA-OFT authors).
• The paper does not name the codebase used for this baseline; openvla-oft is recorded because it is the authors' reference implementation.
• Bimanual YAM setup; single-task fine-tunes averaged 35.5% success over 8 real-world tasks (50 trials each) in that study.

Evidence 2 sources; the cell shows the highest tier

1. T3 · community reportedpapersource 2026-05-04verified 2026-10-24 (1d ago)

   MolmoAct2 paper (Ai2): OpenVLA-OFT is one of four baselines fine-tuned per task and evaluated on a real bimanual YAM setup (8 tasks x 50 trials; OpenVLA-OFT average 35.5%).

   “compare against four strong VLA and world-model baselines: Cosmos Policy (Kim et al., 2026a), X-VLA (Zheng et al., 2025), OpenVLA-OFT (Kim et al., 2025)”

   https://arxiv.org/abs/2605.02881
2. T3 · community reportedpapersource 2026-05-04verified 2026-10-24 (1d ago)

   Appendix C.3 (Real-world Bimanual YAM): every baseline was trained with the same protocol and its converged checkpoint evaluated on the robot.

   “to ensure a fair comparison, we applied the identical training protocol to each of the four baseline policies. All policies were trained to convergence, and the resulting checkpoints were used for evaluation.”

   https://arxiv.org/abs/2605.02881