WP41: final build
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2 changed files +1 −1
modified
paper/main.pdf
+0 −0
Binary file not shown.
modified
paper/sections/ablations.tex
+1 −1
@@ -25,7 +25,7 @@ falsification test passes in the sense intended: shuffling does hurt, so the mod | ||
| 25 | 25 | have much to read. |
| 26 | 26 | |
| 27 | 27 | \paragraph{Fusion, head, regime (H3), identity.} Cross-attention fusion of separate per-scale encoders is slightly |
| 28 | −worse than the shared backbone ($+0.05$\,\%). The Gaussian head is \IfFileExists{tables/ablations_gaussian_note.tex}{\input{tables/ablations_gaussian_note}}{worse than the Student-$t$ head (see Table~\ref{tab:ablations_crps_scaled})}. | |
| 28 | +worse than the shared backbone ($+0.05$\,\%). The Gaussian head is \IfFileExists{tables/ablations_gaussian_note.tex}{\input{tables/ablations_gaussian_note}\unskip.}{worse than the Student-$t$ head (see Table~\ref{tab:ablations_crps_scaled}).} | |
| 29 | 29 | Removing the regime embedding changes nothing ($+0.02$ to $+0.03$\,\%, Holm $p>0.3$ at 30 and 60 minutes), and |
| 30 | 30 | coverage on stress days is identical with and without it (0.57/0.85/0.97 at $H=60$): Hypothesis~\ref{h3} is |
| 31 | 31 | rejected---the global token already carries the regime statistics, and conditioning the decoder on them again |
| 32 | 32 | |