code backup: analysis

analysis/preregistration.md

@@ -93,10 +93,26 @@ All corruption arms run the extended protocol on ground-truth codes (chimera arm
 use both code sources by construction). The solver runs every arm as the zero-cost
 second reasoner.
 
+## Robustness arm (scheduled)
+
+Two single-surface-factor B re-runs on the corruption sample, qwen3.5-4b, extended
+protocol, explicit format: (a) the paraphrased context line
+(harness.B.prompts.PARAPHRASE_PRE_PROMPT), (b) YAML serialization of the identical
+code dict (--serialization yaml; the context line's "JSON" wording swaps to "YAML"
+automatically so no false description rides along). Prediction: B's headline effects
+survive both -- a material change under either is reported as prompt-artifact
+evidence.
+
 ## Analysis plan
 
 Every cross-harness delta: exact question-id intersection (analysis.compare), scene-
-clustered paired bootstrap with fixed seed (analysis.stats), 95% CIs. The primary
+clustered paired bootstrap with fixed seed (analysis.stats), 95% CIs. H21's
+prediction test compares the corruption curve against B RESTRICTED TO THE SAME
+pre-registered sample question ids (a filter over B's full-run records) -- never
+against B's full-benchmark aggregate, whose category mix differs from the balanced
+sample. H6's conflict-question sets are built from the _perturbation.json sidecars
+the single-object runs write per scene (the perturbed object's identity is also
+recomputable from the frozen seed via corruption.run.single_object_info). The primary
 family (H7, H21-H26) is Holm-Bonferroni corrected; supporting and exploratory results
 are reported with CIs but no corrected significance claims. Aggregate category scores
 use only the real, unmodified official VSI-Bench aggregator.