Verify evals on Papers with Code

Hi,

Niels here from the open-source team at Hugging Face.

I've made the following papers and their evaluation results available on Papers with Code:

- [MViTv2: Improved Multiscale Vision Transformers for Classification and Detection](https://paperswithcode.co/paper/2112.01526) — [4 paper-native evaluations](https://paperswithcode.co/paper/2112.01526#results).
- [SlowFast Networks for Video Recognition](https://paperswithcode.co/paper/1812.03982) — [3 paper-native evaluations](https://paperswithcode.co/paper/1812.03982#results).
- [A Multigrid Method for Efficiently Training Video Models](https://paperswithcode.co/paper/1912.00998) — [3 paper-native evaluations](https://paperswithcode.co/paper/1912.00998#results).
- [X3D: Expanding Architectures for Efficient Video Recognition](https://paperswithcode.co/paper/2004.04730) — [3 paper-native evaluations](https://paperswithcode.co/paper/2004.04730#results).

The X3D-XL result currently ranks third on [Charades](https://paperswithcode.co/benchmark/charades?task=video-classification&eval=16865).

The MViTv2-L 312² 40×3 result currently ranks fourth on [Kinetics-700](https://paperswithcode.co/benchmark/kinetics-700?task=video-classification&eval=16861).

The SlowFast 16x8 R101+NL (Kinetics-600 pretrain) result currently ranks fourth on [Charades](https://paperswithcode.co/benchmark/charades?task=video-classification&eval=16907).

Would it be possible to verify these results and let me know if any score, model name, benchmark protocol, or openness metadata should be corrected?

You can also edit the task, methods, project page, and GitHub URL directly from each paper page using your Hugging Face account.

If you'd like to showcase the results in your repository README, you can copy these live leaderboard badges (or use the “Copy PwC badge” button in the Results section):

[](https://paperswithcode.co/api/v1/papers/2004.04730/leaderboard-badge-link?eval=16865)

Kind regards,

Niels

Start by reviewing the four linked Papers with Code paper pages and the Charades and Kinetics-700 benchmark entries. Compare the listed scores, model names, benchmark protocols, and openness metadata with the repository's published results, then report any corrections needed; no repository file or test is identified in the issue.

Written by the indexing model from the issue text.