RoboICL: Embodied In-Context Learning with GPT-6 Astra
Multimodal Models as Few-Shot Robot Learners
General-purpose vision-language models offer a promising way to zero-shot robot control: GPT-6 Astra excels at open-ended and language- or image-conditioned manipulation but remains substantially weaker on high-precision and long-horizon tasks. We introduce RoboICL, an in-context robot-control framework that narrows these gaps without robot-specific parameter updates or a learned VLA. RoboICL separates demonstration context, which provides recorded examples when available, from interaction memory, which accumulates the model’s own actions and observed outcomes. Both use a shared observation → action → receipt → observation grammar. To preserve experience across task stages, RoboICL combines sampled demonstration blocks with bounded anchored memory. Fixed anchors keep earlier rollout interactions available for in-context learning, while the latest interaction supports immediate error correction.
Across 30 RoboDojo tasks, using zero shot for Open and one demonstration elsewhere, RoboICL improves on official zero-shot GPT-6 Astra by 20–27 progress-score points in every category. It leads the leaderboard baselines on Memory and Open, achieves comparable performance to the strongest Precision baseline, and remains competitive on Long-Horizon. Its 30-task Overall score is 50.64, versus 39.56 for the strongest baseline. On a separate ten-task subset, RoboICL scores 60.60, within 2.00 points of the π0.5 + GPT-6 Astra hybrid approach. On three real-robot tasks, mean progress rises from 14.45 at zero shot to 63.33 at one shot and 78.89 at three shots. On two development tasks, optional Jev-gated action reuse reduces GPT-6 Astra calls by 33–48%.
Performance across four task categories in RoboDojo.
RoboDojo category-level comparison.
Choose a category, task, and layout—or click any task row—to inspect its recorded rollout and GPT-6 Astra decisions.
Scroll inside the table to inspect every method →
| Task | RoboICL | RoboProbe | PhysicalRSI | Simate-beta | VPP2-Preview | Liber-0 Preview |
|---|---|---|---|---|---|---|
| Loading benchmark data… | ||||||
Loading recorded decision trace…
Comparison with GPT-as-Policy.
The ten-task RoboDojo panel compares five requested layouts per task with GPT-as-Policy Direct and its reported hybrid approach.
Ten-task RoboDojo comparison.
Scroll horizontally to compare all methods →
| Task | Galaxea G0.5 | OpenWAM α | π0.5 | π0.5 + GPT-6 Astra | GPT-6 Astra Direct |
RoboICL 0-shot |
RoboICL 1-shot |
|---|---|---|---|---|---|---|---|
| Organize the table | 46.33 | 62.50 | 23.33 | 60.00 | 30.00 | 45.00 | 50.00 |
| Classify by language | 1.07 | 1.33 | 0.60 | 38.00 | 60.00 | 44.00 | 44.00† |
| Imitate sorting sequence | 1.67 | 2.90 | 1.60 | 53.00 | 0.00 | 90.00 | 90.00 |
| Arrange largest number | 4.11 | 4.36 | 2.29 | 50.00 | 57.00 | 71.00 | 58.00 |
| Pack objects into a box | 17.12 | 20.83 | 18.36 | 50.00 | 50.00 | 16.00 | 36.00 |
| Classify objects | 10.33 | 5.53 | 24.67 | 71.00 | 100.00 | 100.00 | 71.00 |
| Build Tower | 82.93 | 52.53 | 37.73 | 64.00 | 12.00 | 16.00 | 100.00 |
| Make a Kong in Mahjong | 90.00 | 32.00 | 26.67 | 40.00 | 0.00 | 0.00 | 0.00 |
| Fold clothes | 32.75 | 51.31 | 29.12 | 100.00 | 40.00 | 4.00 | 84.00 |
| Put bottles in a bin | 96.30 | 94.03 | 79.93 | 100.00 | 36.00 | 70.00 | 73.00 |
| Overall | 38.26 | 32.73 | 24.43 | 62.60 | 38.50 | 45.60 | 60.60 |
Citation
@article{liu2026roboicl,
title={RoboICL: Embodied In-Context Learning with GPT-6 Astra},
author={Liu, Fangcheng and Shen, Yeqing and Cheng, Anda and Mi, Weishi and Tang, Chao and Liu, Chenyuan and Xiang, Yushun and Li, Tingguang and Li, Yong-Lu and Tang, Yehui},
journal={arXiv preprint arXiv:2609.34261},
year={2026}
}
