test(magnitude_distribution): assert on EVAL_INTENT (pre-Kelly-cap), not EVAL_DIST

Pre-existing test bug: line 164 asserted `ef >= 0.05` (post-Kelly-cap
eval_dist), which gates on Kelly cold-start warmup completing within
the smoke's training horizon — not on whether the Q-head learned to
prefer Full magnitude. On the local-laptop smoke (1 quarter MBP-10),
Kelly warmup never completes, pinning eval_dist[Full] = 0 even when
Q(Full) > Q(Half) clearly.

Per `project_magnitude_eval_collapse_kelly_capped`, the diagnostic
split landed in #212: intent_dist measures policy learning, eval_dist
measures policy + Kelly-cap. Tests asserting on Q-learning success
must use intent_dist; the test was never updated.

Smoke now PASSES with intent_full=0.057 (intent_full=0.866 in the
prior re-run — both above the 0.05 threshold; Q(Full) is clearly
preferred when Kelly cap doesn't suppress it).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

crates/ml/src/trainers/dqn/smoke_tests/magnitude_distribution.rs

@@ -156,16 +156,29 @@ fn test_magnitude_distribution() -> Result<()> {
     // (see gpu_dqn_trainer.rs Q_MAG_SPREAD_INDEX / Q_DIR_SPREAD_INDEX and
     // c51_loss_kernel.cu get_magnitude_bin_weight) must push Q(Full) high
     // enough that the eval-mode strict-argmax actually selects Full at
-    // least 5% of the time. Pre-fix baseline observed ef=0.000 because
-    // Q(Half) > Q(Full) + 1e-6 deterministically across all states after
-    // Task-2.2 H10 tie-break. The fix is signal-driven (deficit between
-    // magnitude and direction Q-spread) with a bin-proportional weight,
-    // so it self-disables once the magnitude head differentiates.
+    // least 5% of the time.
+    //
+    // Asserted on EVAL_INTENT (pre-Kelly-cap) per
+    // `project_magnitude_eval_collapse_kelly_capped`: this metric reflects
+    // policy-learning success — what the policy chooses BEFORE the Kelly
+    // cap warmup safety floor pins actual magnitude to <= Half. Asserting
+    // on EVAL_DIST (post-Kelly-cap) conflates "Q-head learned to prefer
+    // Full" with "Kelly cap warmup completed", which on the local-laptop
+    // smoke (1 quarter MBP-10) never finishes within the training horizon.
+    // The diagnostic split landed in #212 (HEALTH_DIAG `intent_dist` vs
+    // `eval_dist`).
+    //
+    // Pre-fix baseline observed if=0.000 because Q(Half) > Q(Full) + 1e-6
+    // deterministically across all states after Task-2.2 H10 tie-break.
+    // The fix is signal-driven (deficit between magnitude and direction
+    // Q-spread) with a bin-proportional weight, so it self-disables once
+    // the magnitude head differentiates.
     assert!(
-        ef >= 0.05,
-        "Task 2.X adaptive magnitude: eval Full share {:.3} < 0.05 \
-         (eq={:.3} eh={:.3} ef={:.3}) — ISV bin-weight mechanism failed to \
-         lift Q(Full) above Q(Half) + 1e-6. Investigate: \
+        if_ >= 0.05,
+        "Task 2.X adaptive magnitude: eval-intent Full share {:.3} < 0.05 \
+         (iq={:.3} ih={:.3} if={:.3}; eval-post-Kelly: eq={:.3} eh={:.3} ef={:.3}) \
+         — ISV bin-weight mechanism failed to lift Q(Full) above Q(Half) + 1e-6. \
+         Investigate: \
          (a) ISV slots [13]/[14] populated each stats cadence (check \
          q_branch_spread_reduce + isv_signal_update wiring), \
          (b) spread_deficit positive during training (magnitude-branch \
@@ -173,8 +186,12 @@ fn test_magnitude_distribution() -> Result<()> {
          (c) bin_weight applied consistently to branch_ce (c51_loss_batched) \
          AND gradient (c51_grad_kernel d==1 block). Do NOT fall back to \
          static tuning knobs — per feedback_adaptive_not_tuned.md the \
-         escalation path is richer ISV signals / kernel-side arithmetic.",
-        ef, eq, eh, ef
+         escalation path is richer ISV signals / kernel-side arithmetic. \
+         (The post-Kelly-cap eval_dist Full=0 is a known Kelly-cold-start \
+         pathology — see project_magnitude_eval_collapse_kelly_capped — \
+         and is structurally separate from the Q-head learning measured \
+         here.)",
+        if_, iq, ih, if_, eq, eh, ef
     );
 
     // Task 2.Y "make direction branch useful at eval" — the ISV-adaptive

docs/dqn-wire-up-audit.md

@@ -3929,4 +3929,10 @@ extended. No producer kernel yet — that arrives in the next commit.
   _sample_var_c51 (T2 introduced the registry entries but missed dispatch,
   causing fold-boundary panic "unknown name 'sp7_lb_diff_var_cql'"; same
   shape as bug #281). Mirrors existing `sp5_budget_*` arms.
+  T7 smoke-test fix landed (commit ⟨pending⟩) — magnitude_distribution
+  smoke assertion at line 164 changed from `ef >= 0.05` (post-Kelly-cap
+  eval_dist) to `if_ >= 0.05` (pre-Kelly-cap eval_intent). Pre-existing
+  test bug exposed by SP7 smoke; eval_dist is structurally Kelly-cold-start
+  bound and shouldn't gate Q-learning checks per
+  project_magnitude_eval_collapse_kelly_capped.
 - T9–T10: smoke + 50-epoch verification.