FE-1549: Rotate sweep parameter draws by a seed-derived shift

[kube]

🌟 What is the purpose of this PR?

Range sweeps drew each run's parameter values from a bare per-axis Halton sequence: every experiment explored the identical value sequence regardless of its seed, and deterministic low-discrepancy points are not exchangeable random samples. This rotates each axis's draws by a seed-derived shift (Cranley–Patterson), making them unbiased over the seed while keeping the low-discrepancy convergence, the even early coverage, and prefix stability. Stacked on #9419.

🔗 Related links

• FE-1549 (internal)
• Base PR: FE-1546: Stream range sweeps as one stochastic simulation #9419

🔍 What does this change?

• sweepRunFraction takes the experiment seed and rotates the radical-inverse fraction by a per-axis shift hashed from the seed and the axis index (32-bit Math.imul hash), modulo 1.
• sweepRangeRuns threads the seed through; the sweep session passes its experiment seed. A draw depends only on the seed, the axis, and the global run index, so ladder batches still extend the same sequence and the selection cache stays valid.
• Tests: prefix stability, unit-interval spread across seeds, per-axis distinctness, exact reproduction for one seed, distinct sequences across seeds.
• Docs: the user-guide sweep section and the parameter-sweeps architecture page describe the seed-shifted draws.

Pre-Merge Checklist 🚀

🚢 Has this modified a publishable library?

This PR:

• modifies an npm-publishable library and I have added a changeset file(s)

📜 Does this require a change to the docs?

The changes in this PR:

• require changes to docs which are made as part of this PR

🕸️ Does this require a change to the Turbo Graph?

The changes in this PR:

• do not affect the execution graph

🛡 What tests cover this?

• parameter-grid.test.ts (sweepRunFraction suite, extended) and the existing sweep-session.test.ts range-draw assertions, which now exercise the seeded path.

❓ How to test this?

1. Run yarn workspace @hashintel/petrinaut dev and open "Simulate / SweepNavigator / Real compute on CPU".
2. Select a range: the streamed distribution covers the interval from the first rung, as before.
3. Create two sweeps with different seeds over the same range: their per-run draws differ; re-running the same seed reproduces them.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
hash	Ready	Preview	Aug 29, 2026 2:51am