neural: real-MNIST external-validity tier (collapse + grounding)

Confirms model collapse and its arrest by grounding on REAL images, not
just the synthetic sandbox. A conv VAE (the canonical generative-collapse
model) is retrained each generation on its own generated digits, with a
fraction g of fresh real MNIST mixed in. Modes = digit class x stroke-
thickness bin (K=30, Zipf, ~18 tail modes); the oracle is a frozen CNN +
deterministic thickness at 98.5% mode accuracy (30x30 confusion matrix
recorded in the manifest as the measurement-noise floor).

Result (4 reps): dry (g=0) collapses to a single mode -- forward-KL
0.5->18, support 30->1, tail 1.0->0.06, H->0 -- while 10% grounding holds
all 30 modes (KL~0.6, full tail, H~0.9). Signs, not magnitudes (blueprint
3.5); the exact synthetic oracle stays the quantitative anchor. The VAE
needs ~10% grounding vs the synthetic histogram's ~5%, consistent with the
grounding finding that trained nets need more than the exact operator.

Plugs into the existing data-agnostic contract (metrics/grounding/output
reused verbatim): mnist_data (thickness bins, class x thickness bijection,
MnistSampler), mnist_oracle (ClassifierOracle + confusion matrix),
mnist_vae (ConvVAEGenerator), mnist_loop (run_mnist_lineage), kind=
mnist_lineage dispatch, MnistCfg/OracleCfg. Figures: plot_mnist (parquet-
only) + mnist_montage (eyeball diagnostic showing digits degenerate to one
blurry mode). make mnist / make env-mnist, kept out of the make neural
loop. 99 tests green (+5 torchvision-gated).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

results/mnist_collapse/resolved_config.yaml

@@ -0,0 +1,110 @@
+experiment: mnist_collapse
+seed: 20260705
+n_replicates: 4
+source_config:
+  experiment: mnist_collapse
+  kind: mnist_lineage
+  seed: 20260705
+  n_replicates: 4
+  mnist:
+    K: 30
+    n_classes: 10
+    style_bins: 3
+    R: 1
+    tail: zipf
+    zipf_s: 1.5
+    tail_threshold: 0.01
+    init: truth
+    data_root: data
+  model:
+    kind: convvae
+    latent: 32
+    epochs: 30
+    lr: 0.001
+    batch_size: 256
+    beta: 1.0
+    n_eval: 10000
+  oracle:
+    epochs: 5
+    lr: 0.001
+    batch_size: 256
+    cache: models/mnist_cnn.pt
+  dynamics:
+    n: 6000
+    grounding:
+      m: 0
+      policy: proportional
+  generations: 15
+  metrics:
+    kl_floor: 1.0e-09
+    support_eps: 1.0e-09
+  sweep:
+  - param: g
+    values:
+    - 0.0
+    - 0.1
+  output:
+    dir: results/mnist_collapse
+grid:
+- label:
+    g: 0.0
+    m: 0
+  lineage_cfg:
+    mnist:
+      K: 30
+      n_classes: 10
+      style_bins: 3
+      R: 1
+      tail: zipf
+      zipf_s: 1.5
+      tail_threshold: 0.01
+      init: truth
+      data_root: data
+    model:
+      kind: convvae
+      latent: 32
+      epochs: 30
+      lr: 0.001
+      batch_size: 256
+      beta: 1.0
+      n_eval: 10000
+    dynamics:
+      n: 6000
+      grounding:
+        m: 0
+        policy: proportional
+    generations: 15
+    metrics:
+      kl_floor: 1.0e-09
+      support_eps: 1.0e-09
+- label:
+    g: 0.1
+    m: 667
+  lineage_cfg:
+    mnist:
+      K: 30
+      n_classes: 10
+      style_bins: 3
+      R: 1
+      tail: zipf
+      zipf_s: 1.5
+      tail_threshold: 0.01
+      init: truth
+      data_root: data
+    model:
+      kind: convvae
+      latent: 32
+      epochs: 30
+      lr: 0.001
+      batch_size: 256
+      beta: 1.0
+      n_eval: 10000
+    dynamics:
+      n: 6000
+      grounding:
+        m: 667
+        policy: proportional
+    generations: 15
+    metrics:
+      kl_floor: 1.0e-09
+      support_eps: 1.0e-09