bench(worldmodel): the fourth condition is rank preservation, not accuracy

[aurascoper]

Node 23's adjudicating sweep: inject latent error into an otherwise-perfect rollout and measure control against magnitude.

Reading (a) is refuted — control degrades gracefully

noise sd	equiv MSE	error / goal signal	success
0.000	0	0.00	80.5%
0.040	0.00160	0.43	81.0%
0.074	0.00548	0.79	77.0% ← multi-step trained level
0.100	0.01000	1.07	72.0%
0.152	0.02311	1.62	65.0% ← incumbent level
0.250	0.06250	2.67	59.0%

Median latent goal signal is 0.5303, so the incumbent's rollout error is 1.62× the entire goal signal — and random error at that magnitude still yields 65.0%, where the real predictor yields 35.0%.

Magnitude is not the explanation. Even error 2.67× the goal signal leaves control at 59%.

The fourth condition is rank preservation

120 episodes × 64 candidates:

ranking correlation	Spearman (median)
predictor cost ↔ true latent cost	+0.389
predictor cost ↔ true position cost	−0.018

The predictor ranks candidates acceptably in latent space, and its ranking is uncorrelated with the objective that decides success. Only 60.8% of episodes reach even ρ > 0.3 against the latent cost.

Why injected noise is benign and the predictor is not

Random noise perturbs each candidate independently, so the ordering survives in expectation and the CEM's elite set stays informative. A predictor whose error is structured — concentrated in the 31 latent directions that don't encode position, while the one that does is swamped — destroys the ordering at any MSE.

The two failure modes are not comparable by magnitude. That is exactly why five nodes of accuracy-improving repairs changed nothing.

Verdict

Reject "improve the forward model's accuracy" as a route to control, and with it the framing driving nodes 19–23. MSE over a 32-d latent rewards average accuracy across dimensions dominated by the frozen factors; the planner needs correct ordering along the single direction encoding position.

65% from noisy-but-rank-preserving versus 35% from accurate-but-rank-destroying, at the same error magnitude.

This vindicates node 13

Node 13 named "a control-aware goal metric" as one of two fixes and explicitly reserved it as a design decision for the user, not an autonomous build.

Nodes 14–23 pursued the other branch — representation, benchmark, frame, action channel, forward model. Five real defects found, four fixed, control moved by nothing. The measured reason is now on record.

Next

A design decision, and per the node-13 stop rule it belongs to you, not to an autonomous build. The measured requirement is a training signal that optimises ranking with respect to the control objective rather than reconstruction error in latent space.

One bounded check should precede it: measure Spearman(true latent cost, true position cost) directly. Node 22 showed the true latent cost yields 76%, so it should be substantial — and if it is not, the goal metric is the defect rather than the predictor's use of it.

Ledger-only. Verified: smoke 2 runs; ledger and ADR gates green.