Skip to content

docs: publish Wave Forcing 4-step quality results #1

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Icyw-git wants to merge 1 commit into
base: main
Choose a base branch
from
Open

docs: publish Wave Forcing 4-step quality results #1

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 14 additions & 8 deletions page/index.html
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -732,29 +732,29 @@ <h2>Acceleration doesn't trade<br>away quality.</h2>
<tbody>
<tr>
<td><b class="quality-metric">Rolling Forcing<sup>*</sup></b></td>
<td><div class="quality-cell best">97.78</div></td>
<td><div class="quality-cell">97.78</div></td>
<td><div class="quality-cell">97.20</div></td>
<td><div class="quality-cell">96.23</div></td>
<td><div class="quality-cell best">98.84</div></td>
<td><div class="quality-cell">53.62</div></td>
<td><div class="quality-cell best">69.62</div></td>
<td><div class="quality-cell best">1.111</div></td>
<td><div class="quality-cell">69.62</div></td>
<td><div class="quality-cell">1.111</div></td>
</tr>
<tr>
<td><b class="quality-metric">Rolling Forcing on WaveRT<sup>*</sup></b></td>
<td><div class="quality-cell">97.36</div></td>
<td><div class="quality-cell">96.69</div></td>
<td><div class="quality-cell">96.11</div></td>
<td><div class="quality-cell">98.50</div></td>
<td><div class="quality-cell best">54.52</div></td>
<td><div class="quality-cell">54.52</div></td>
<td><div class="quality-cell">66.32</div></td>
<td><div class="quality-cell">2.680</div></td>
</tr>
<tr>
<td class="wave-col"><b class="quality-metric">Wave Forcing<sup>*</sup></b></td>
<td class="wave-col"><div class="quality-cell">97.55</div></td>
<td class="wave-col"><div class="quality-cell best">97.66</div></td>
<td class="wave-col"><div class="quality-cell best">96.69</div></td>
<td class="wave-col"><div class="quality-cell">97.66</div></td>
<td class="wave-col"><div class="quality-cell">96.69</div></td>
<td class="wave-col"><div class="quality-cell">98.28</div></td>
<td class="wave-col"><div class="quality-cell">54.31</div></td>
<td class="wave-col"><div class="quality-cell">68.89</div></td>
Expand All @@ -772,7 +772,13 @@ <h2>Acceleration doesn't trade<br>away quality.</h2>
</tr>
<tr>
<td class="wave-col"><b class="quality-metric">Wave Forcing 4-step</b></td>
<td class="pending-col" colspan="7"><div class="quality-pending"><b>Stay tuned</b><small>Wave Forcing 4-step · results will be posted here</small></div></td>
<td><div class="quality-cell best">98.08</div></td>
<td><div class="quality-cell best">97.99</div></td>
<td><div class="quality-cell best">96.70</div></td>
<td><div class="quality-cell">98.76</div></td>
<td><div class="quality-cell best">58.69</div></td>
<td><div class="quality-cell best">73.27</div></td>
<td><div class="quality-cell best">0.319</div></td>
</tr>
</tbody>
</table>
Expand All @@ -796,7 +802,7 @@ <h2>Acceleration doesn't trade<br>away quality.</h2>
</div>
</div>

<div class="caveat reveal"><span>Protocol</span><p>The 5-step table runs all three arms on the same 50 MovieGen prompts, one 1593-frame video per prompt at 16 FPS / 99.56 s. Rolling Forcing on WaveRT is the same Rolling Forcing checkpoint decoded with Wave Forcing's block-causal inference. The 4-step table forces the same three 5-step checkpoints to a 4-step schedule and adds Causal Forcing (Long), which runs its native 4-step long-video inference; Wave Forcing 4-step results are pending and will be posted as the sweep completes. The algorithm × runtime table uses 10 aligned prompts for VBench-Long quality; FPS is DiT-only throughput from a one-shot 399-latent-frame run. Quality Drift is computed following the Rolling Forcing paper, lower-is-better; every other quality metric is higher-is-better.</p></div>
<div class="caveat reveal"><span>Protocol</span><p>The 5-step table runs all three arms on the same 50 MovieGen prompts, one 1593-frame video per prompt at 16 FPS / 99.56 s. Rolling Forcing on WaveRT is the same Rolling Forcing checkpoint decoded with Wave Forcing's block-causal inference. The 4-step table forces the same three 5-step checkpoints to a 4-step schedule and adds Causal Forcing (Long), which runs its native 4-step long-video inference; Wave Forcing 4-step results are included. The algorithm × runtime table uses 10 aligned prompts for VBench-Long quality; FPS is DiT-only throughput from a one-shot 399-latent-frame run. Quality Drift is computed following the Rolling Forcing paper, lower-is-better; every other quality metric is higher-is-better.</p></div>
</section>

<section class="closing">
Expand Down