Princeton's Action Self-Distillation Teaches Robots Language Navigation with Zero Human Labels
shahdhruv_ · x · 2026-10-23
Princeton researchers introduce Action Self-Distillation (ASD, CoRL 2026), where a single open-source VLM acts as both teacher and student to learn language goal-conditioned reasoning and navigation directly from robot demonstrations—no human-annotated goals or reasoning traces needed.
- The teacher infers goals and reasoning from demonstrated trajectories; the student learns to reproduce them from camera images and commands
- Trained on 100 hours of navigation data vs. 9,500 hours for OmniVLA
- Zero-shot on 13 real-world tasks: 95% outdoor success, +30 points indoors over OmniVLA
- Self-distillation hits 80% on ablation tasks vs. 36% for SFT on Claude Opus 5 labels
More from Embodied
- Embodied First Exam benchmarks coding agents on 1,090 tasks across 44 robots — ManlingLi_ · 2026-10-23
- Dev builds ESP32 desk display showing Claude Code and Codex agent status via local Go helper — tonylook · 2026-10-23
- Dream Labs emerges from stealth with Dream-1 robot foundation model at $3/hour — scott_e_reed · 2026-10-23
- AdaDex turns seconds of human hand motion into generalizable dexterous robot skills — mangahomanga · 2026-10-23
- Perplexity's Portable Computer lands on AMD Ryzen AI Max, running local 27B model — AravSrinivas · 2026-10-23
- MIT CSAIL's X-Hinges 3D-Prints Everyday Objects With Built-In Sensing Across Three Motion Axes — MIT News AI · 2026-10-23