Publication visuals
Publication visuals
Images accompany papers authored by Pengxiang Ding. Previews are locally hosted WebP derivatives of the original figures or unaltered video frames. Full images open on demand. Captions and source links also appear in the website image viewer.
Sources
| Work | Original source | Capture | Rights notice |
|---|---|---|---|
| ceed-vla | Original · Context | CEED-VLA method overview · Figure 2, page 4. PDF rectangle in points: (135, 115, 480, 287). | Source authors retain rights |
| hif-vla | Original · Context | Real-world block-placement demo · Video frame at 00:06. | CC BY-SA 4.0 (project page) |
| cubic | Original · Context | CUBic overall framework · Figure 1, page 2. PDF rectangle in points: (58, 72, 552, 255). | Source authors retain rights |
| ud-vla | Original · Context | Joint Discrete Denoising Diffusion Process (JD3P) · Method overview. | Source authors retain rights |
| trajbooster | Original · Context | Unitree G1 deployment demo · Video frame at 00:05. | Source authors retain rights |
| vla-adapter | Original · Context | VLA-Adapter and Bridge Attention · Method overview. | CC BY-SA 4.0 (project page) |
| carp | Original · Context | Coarse-to-fine action generation · Comparison with autoregressive and diffusion policies. | Source authors retain rights |
| quart-online | Original · Context | QUART-Online · Action chunk discretization for quadruped control. | Source authors retain rights |
| quar-vla | Original · Context | QUAR-VLA training and deployment · Figure 2, page 3. PDF rectangle in points: (134, 114, 481, 230). | Source authors retain rights |
Processing
- PDF captures isolate the complete referenced figure without its surrounding body text.
- Video captures use the stated timestamp; the screenshots do not assert that the surrounding video is unedited.
- Images are proportionally resized and encoded as WebP. Diagram previews use contain, preserving every panel. Desktop demo thumbnails may crop the display; the image viewer shows the full frame.
- The HiF-VLA and VLA-Adapter project pages declare CC BY-SA 4.0. Derivative image files for those projects are shared under the same license.
- Public source URLs and captions are recorded in
_data/home_publication_media.json.
Link correction
CARP previously linked to arXiv 2506.13725, which is CEED-VLA. Its paper link now points to arXiv 2412.06782, as verified against the CARP project page.