KAIST Humanoid v0.7: version history

Current record →

Nothing is deleted from the index. Every change to this record is kept: a full copy of each version since 6 Sep 2026, 05:24 UTC, and the field-by-field updates applied before that. See every robot from KAIST Dynamic Robot Control & Design Lab.

Versions · 3 recorded

Version 370 · 18 Sep 2026, 02:00 UTC · score then 57.5
- ai model: Deep Reinforcement Learning (DRL) with human motion demonstration priors and Sim2Real transfer → Deep Reinforcement Learning (proprioceptive blind locomotion policies)
Changed by a direct correction or a scoring update, not a research proposal.
Version 296 · 15 Sep 2026, 04:28 UTC · score then 57.5

No field changed; the score was recalculated.

Changed by: Update KAIST Humanoid v0.7 (confirmed: ai_model) (research agent)
First recorded version · 6 Sep 2026, 05:24 UTC · score then 57.5

The record as it stood when full copies began to be kept.

Earlier changes · 1, before full copies were kept

Update KAIST Humanoid v0.7 · 2026-08-23 01:04 UTC · research agent
- autonomy level: research → task-autonomous
- ai model: RL locomotion policies trained in simulation, transferred to hardware → Deep Reinforcement Learning (DRL) with human motion demonstration priors and Sim2Real transfer
- speed (m/s): 3.6 → 3.61
- dexterous hands: — → False