[dionhaefner]

Relevant issue or PR

Stacked on #203 (docs redesign), into which it merges. It presents the experiments added in #272 and #315. Until those are merged it changes nothing visible, because a setup without results is left off the scoreboard.

Description of changes

• Scoreboard: a new ns-grid setup, Forced periodic box (Kolmogorov flow).
  • It claims forward/kolmogorov_laminar, forward/kolmogorov and gradient/chain_fd_check, so they don't fall into the periodic setup.
  • Its forward gates are the two forward checks, and its gradient gate is the chained check. It has no optimization task.
• Protocol (index.qmd): forward accuracy now names the Kolmogorov flow. It says the turbulent case is chaotic, so a solver must converge towards the reference on a ladder of grids at observed order 0.8 or more, instead of meeting a fixed error bound. A new paragraph under gradient accuracy explains the chained check. Training through a solver passes each call a tiny cotangent, and the check catches VJPs that are only accurate for large ones.
• Results page: readable titles for the three experiments. The rest of that section comes from the experiments themselves: their plot descriptions and plots.

Testing done

• tests/test_docs_generate.py passes.
• The pages were generated from a local A100 run of the earlier version of this branch, which presented the solver-in-the-loop check. They have not been regenerated since the chained check replaced it.

[PasteurBot]

📊 View the full benchmark results

The rendered docs preview has every plot for this run (forward accuracy, gradients, cost, optimization) merged with existing baseline results on main. The summary below reports pass/fail status.

---

Mosaic status

Benchmarks: not applicable (no mosaic/ changes).

[dionhaefner]

Closing to keep the Kolmogorov work in one PR (#272). Its three-file change (scoreboard setup, protocol paragraphs, result titles) stays on doc/kolmogorov-results and gets added to #272 once #203 lands, since the scoreboard only exists there.