R2RI
Robot-to-Robot Interaction (R2RI)
A large-scale synthetic multimodal dataset with >6.5M frames and ~5000 videos for evaluating robot-robot interaction perception tasks.
How to run
$ npm i -g lettokenburn-cli$ ltb bench info r2ri
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
7 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| DINO | 55.4 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| RT-DETR | 44.8 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| Faster R-CNN | 29.6 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| Def. DETR | 24.8 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| YOLO11l | 21.2 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| YOLO26l | 20.1 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| DETR | 19.1 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsAbout
R2RI is a synthetic dataset featuring humanoid robots in realistic social interactions, captured from egocentric and exocentric views using synchronized RGB and event cameras. It supports eight perception tasks including detection, pose estimation, and action recognition.
The benchmark evaluates state-of-the-art models across RGB and event modalities, providing baseline performance metrics to analyze modality-specific strengths and trade-offs in robot interaction scenarios.