v1.2.0: leave-one-family-out routing, corrected BH-FDR statistics, paper-cited result sets missing from 1.1.0

Adds routing_family_holdout and meta_router_family_holdout; regenerates paper_c/ablation_statistics over 9 PDEs (27 tests) and statistical_tests.json with a correct Benjamini-Hochberg step-up; adds hypino_matched_budget*, pretrain_budget_sweep, clamping_ablation, noisy_data_probe*, rd_speed_D*, multiscale2d_policy_suite_a/b; corrects README, CLAIM_INVENTORY and croissant (1.2.0).

CLAIM_INVENTORY.md

@@ -1,5 +1,8 @@
 # Claim Inventory — docs/paper_combined/main.tex
 
+> **Superseded (2026-10-28).** This inventory was written for the submission and several rows are wrong: #1 (1,539 runs does not reproduce; the release holds 3,175 recorded runs), #4–5 (causal × pretraining is +5.9% with 6/9 significant and −2.4% without, not 7/9 and −4.2%), #13 (0.97 → 0.98), #21–22, and the family-macro claim, which came from leave-one-PDE-out folds rather than a family holdout. The authoritative, executable list is `scripts/verify_claims.py` in the source repository.
+
+
 Tracks every headline number (abstract / intro / key section claims / conclusion) against the authoritative artifact. Any mismatch is fixed by edits to `main.tex`; no artifacts are modified.
 
 | # | Claim (paper) | Artifact | Computed value | Paper value | Status | Fix |
@@ -23,6 +26,9 @@ Tracks every headline number (abstract / intro / key section claims / conclusion
 | 17 | HyPINO Heat1D adapter is executable but out-of-distribution and not same-budget comparable | `paper_combined/hypino_heat1d_adapter/results.json` | Rel. L2 2.412, MSE 0.153, MAE 0.352 | reported in appendix adapter table | match | added to document direct-adapter attempt rather than omit failed comparability |
 | 18 | HyPINO-generated target PINN after 500-step adaptation is competitive on Heat1D and AdvDiff1D | `paper_combined/hypino_adaptation_2pde/results.json` | Heat1D Rel. L2 0.0115 ± 0.0005; AdvDiff1D Rel. L2 0.0043 ± 0.0001 | reported in appendix adapter table | match | 2-PDE, 3-seed same-task adaptation probe, not full multi-PDE reproduction |
 | 19 | PINNacle 5-task executable subset run completes | `paper_combined/pinnacle_subset_5task/results.json` | Burgers1D/Wave1D/Poisson1D/Helmholtz2D/Heat2D-Multiscale at 200 iterations; mean Rel. L2 1.090 ± 0.395 | reported in appendix PINNacle subset table | match | smoke suite for benchmark compatibility |
+| 20 | HyPINO matched-budget: hybrid loses, evidence for pretraining-as-dominance | `experiments/results/paper_combined/hypino_matched_budget/aggregate.json` | 4 PDEs × 10 seeds. Heat1D HyPINO 0.0108 vs Hybrid 0.2163 (~20×); AdvDiff1D HyPINO 0.0047 vs Hybrid 0.4277 (~90×); AllenCahn1D PINN-fallback 0.1681 vs Hybrid 0.3506; Burgers1D NaN. | reported in appendix `tab:hypino-matched` | match | from-scratch hybrid loses under matched wall-budget; framing: pretraining is the dominant factor |
+| 21 | Pretraining-budget sweep: causal benefit conditional on ≥200 epochs, not monotonic | `experiments/results/paper_combined/pretrain_budget_sweep/aggregate.json` | 3 PDEs × 4 budgets × 2 conditions × 10 seeds = 240 runs. Causal Δ% ≤ +5.3% peak; AdvDiff1D budget=400 +5.3%; not monotonic across budgets. | reported in appendix `tab:pretrain-sweep` | match | reproduces +5.9% headline near budget=400 on AdvDiff1D; noise-dominated below 200 epochs |
+| 22 | Clamping ablation null: weight floor 0.1 had no effect | `experiments/results/paper_combined/clamping_ablation/aggregate.json` | 9 PDEs × 10 seeds. Empirical `final_min_weight` ranges 0.94 (AllenCahn) to 0.997 (Heat1D). Unclamped ≡ clamped to 4 decimal places on rel-L2 and PDE-residual. | reported in appendix `tab:clamping-null` | match | weight collapse not directly observed at 200/500/200, ε=1.0; mechanism narrative retained as hypothesis |
 
 ## Authoritative per-k regret-bound breakdown (canonical)
 

README.md

@@ -39,7 +39,7 @@ A controlled benchmark for evaluating training-policy selection in hybrid Physic
 
 ## What is in this repository
 
-This Hugging Face dataset hosts the **benchmark protocol artifacts** + **logged results** of 1,539 controlled training runs on 13 PDE configurations spanning 8 equation families. The dataset is **not raw simulation data** (PDE reference solutions are analytical and regenerated at runtime); it is the **decision-relevant evaluation record** that supports the paper's headline claims.
+This Hugging Face dataset hosts the **benchmark protocol artifacts** + **logged results** of 3,175 recorded training runs on 13 PDE configurations spanning 8 equation families. The dataset is **not raw simulation data** (PDE reference solutions are analytical and regenerated at runtime); it is the **decision-relevant evaluation record** that supports the paper's headline claims.
 
 ```
 results_paper_combined/
@@ -52,6 +52,7 @@ results_paper_combined/
 │   └── results.json              # PDE-aware router (84.6% with 5+7 features)
 ├── stage_ablation/
 │   └── *_ablation.json           # 9 PDEs × 2×2 (Stage1, Stage3) factorial × 10 seeds
+│   # statistics over all 27 tests (BH-FDR step-up, corrected in v1.2.0): results/paper_c/ablation_statistics/
 ├── adaptive_epsilon/
 │   └── results.json              # 4 PDEs × 4 conditions × 5 seeds (causal-weight collapse)
 ├── wang_comparison/
@@ -59,14 +60,22 @@ results_paper_combined/
 ├── hypino_*/                     # HyPINO native, adapter, target-PINN adaptation probes
 ├── pinnacle_subset_5task/
 │   └── results.json              # PINNacle 5-task executable subset
-├── statistical_tests.json        # BH-FDR corrected, 27 tests
+├── statistical_tests.json        # BH-FDR corrected, 13 core tests
 │   # --- added in v1.1.0 (camera-ready) ---
 ├── noise_sparsity_sweep*/        # 6 PDEs x 3 noise x 4 data sizes x 2 conditions x 5 seeds (720 runs)
 ├── heat3d_probe/                 # d=3 pretraining contrast, 10 seeds
 ├── multiscale2d_probe/           # MultiscaleHeat2D contrast + per-mode projection, 10 seeds
 ├── multiscale2d_policy_suite/    # full 5-policy library on MultiscaleHeat2D, 50 runs
 ├── dissociation_transfer/        # accuracy-regret dissociation on tabular classifier selection
-└── k20_gap_check.json            # regret-bound check at probe window k=20
+├── k20_gap_check.json            # regret-bound check at probe window k=20
+│   # --- added in v1.2.0 ---
+├── routing_family_holdout/       # leave-one-family-out routing, base routers (whole family removed from training)
+├── meta_router_family_holdout/   # leave-one-family-out routing, PDE-aware routers
+├── hypino_matched_budget*/       # HyPINO vs hybrid at matched wall-clock budget (paper-cited, missing from 1.1.0)
+├── pretrain_budget_sweep/        # Stage-1 budget sweep, 240 runs (paper-cited, missing from 1.1.0)
+├── clamping_ablation/            # causal-weight clamping null result (paper-cited, missing from 1.1.0)
+├── noisy_data_probe*/            # rebuttal-period Heat1D noise probes
+└── rd_speed_D*/                  # reaction-diffusion front-speed sweeps
 ```
 
 ## How to load
@@ -94,13 +103,14 @@ Every headline claim in the paper is recomputed from these JSONs by the script `
 
 | Claim | Source artifact | Computed | Paper |
 |---|---|---|---|
-| Total runs | union | 1,510 enumerable + 29 probes | 1,539 |
+| Total recorded runs | all JSONs, one count per (training_time, seed) | 450 routing + 360 ablation + 449 single-PDE + 706 diagnostic + 1,210 robustness | 3,175 |
 | Nested PDE-aware routing accuracy | `meta_router/results.json` | 84.62% (11/13) | 84.6% |
 | Full RF+PDE routing | `meta_router/results.json` | 61.54% (8/13) | 61.5% |
 | Stage-2 compute savings | analytical | 60.0% (K=3, Eₚ=50, E₂=500) | 60% |
 | Diagnostic regret bound | `regret_bound_validation.json` | 33/52 overall, 33/39 if k∈{10,50,100} | 33/52, 33/39 |
 | Accuracy–regret dissociation | `routing_evaluation/results.json` folds + `results/paper_a/*` runs | 0.0850 vs 0.0011 (77.6× unrounded) | 77× |
-| Family-macro accuracy (LOPO-F) | `routing_evaluation/results.json` | 81.2% physics-loss-final (6/8) | 81.2% |
+| Leave-one-family-out routing | `routing_family_holdout/`, `meta_router_family_holdout/` | PDE-aware GBR+PDE 84.6% → 69.2%, RF+PDE 84.6% → 61.5%; physics-loss heuristic 69.2% (family macro 81.2%) | same |
+| Causal loss × pretraining | `paper_c/ablation_statistics/results.json` (27 BH-FDR tests) | +5.9% with (9/9 positive, 6 significant); −2.4% without (5/9 negative, 3 significant) | same |
 | KdV1D causal effect (PDE residual) | `rq1e_kdv_10seed_causal/results.json` | 0.0096 / 0.0105 | 0.010 / 0.011 |
 | Adaptive-ε min causal weight (AdvDiff1D) | `adaptive_epsilon/results.json` | 0.78 (fixed, pretrained) → 0.98 (adaptive, no pretraining) | same |
 
@@ -112,7 +122,7 @@ Every headline claim in the paper is recomputed from these JSONs by the script `
 
 ## Croissant metadata
 
-`croissant.json` (Croissant 1.1, version 1.1.0) carries 13 `rai:` fields. `mlcroissant validate --jsonld croissant.json` (mlcroissant 1.1.0) reports 0 errors and 1 warning: the `@context` includes `examples`, `equivalentProperty` and `samplingRate`, which are Croissant 1.1 keys absent from the 1.0 reference context.
+`croissant.json` (Croissant 1.1, version 1.2.0) carries 13 `rai:` fields. `mlcroissant validate --jsonld croissant.json` (mlcroissant 1.1.0) reports 0 errors and 1 warning: our `@context` lacks three keys of the reference context (`examples`, `equivalentProperty`, `samplingRate`), none of which the file uses.
 
 ## Citation
 

croissant.json

@@ -11122,10 +11122,10 @@
   ],
   "license": "https://choosealicense.com/licenses/apache-2.0/",
   "url": "https://huggingface.co/datasets/suanlab/PINNBench",
-  "version": "1.1.0",
+  "version": "1.2.0",
   "datePublished": "2026-05-04",
   "citeAs": "Suan Lee, Namhyeon Kim, and Dongmin Jin. PINNBench: A Benchmark and Evaluation Study of Training Policy Selection in Hybrid PINN-Operator Solvers. In Advances in Neural Information Processing Systems (NeurIPS), Evaluations and Datasets Track, 2026.",
-  "rai:dataLimitations": "The benchmark suite contains 13 PDE configurations across 8 equation families and is dominated by 1D problems with a single 2D Wave equation and a single 2D Navier-Stokes configuration. Headline results use analytical, noise-free Stage-1 reference data. Version 1.1.0 adds controlled corruption (relative Gaussian noise up to 20%) and sparsification (down to 16 reference points), plus a d=3 and a multi-scale configuration; real-instrument data and inverse problems remain unmeasured, and on the multi-scale problem no policy recovers the fine scale. All training is performed for a fixed compute budget on a single 4xNVIDIA TITAN Xp node; results may not generalise to substantially larger architectures or different optimiser families. Routing meta-features include hand-coded PDE structural attributes that may partly proxy PDE family identity in this small suite, although the family-macro physics-loss-final result (81.2 percent, no PDE identity features used) partially mitigates this concern.",
+  "rai:dataLimitations": "The benchmark suite contains 13 PDE configurations across 8 equation families and is dominated by 1D problems with a single 2D Wave equation and a single 2D Navier-Stokes configuration. Headline results use analytical, noise-free Stage-1 reference data. Versions 1.1.0 and 1.2.0 add controlled corruption (relative Gaussian noise up to 20%) and sparsification (down to 16 reference points), plus a d=3 and a multi-scale configuration; real-instrument data and inverse problems remain unmeasured, and on the multi-scale problem no policy recovers the fine scale. All training is performed for a fixed compute budget on a single 4xNVIDIA TITAN Xp node; results may not generalise to substantially larger architectures or different optimiser families. Routing meta-features include hand-coded PDE structural attributes that may partly proxy PDE family identity in this small suite, and a real leave-one-family-out evaluation confirms it partly does: PDE-aware routing falls from 84.6 to 61.5-69.2 percent when the held-out PDE's whole family is removed from training, level with a training-free physics-loss heuristic (69.2 percent).",
   "rai:dataBiases": "By construction the benchmark draws PDEs from diffusion, advection, nonlinear waves, dispersive, reaction-diffusion, and Navier-Stokes regimes to reduce family-specific bias, but: (a) it over-represents 1D problems; (b) reference solutions are analytical exact-solutions, which biases methods that exploit smooth ground truth; (c) the routing benchmark records baseline-policy decisions for a fixed library of five training policies, which biases evaluation toward routing methods that generalise within this policy set. We deliberately publish negative and conditional findings to expose these biases rather than hide them.",
   "rai:personalSensitiveInformation": "None. The benchmark consists of synthetic numerical-PDE simulations with analytical reference solutions. No personal, identifying, biometric, behavioural, location, health, or other sensitive data is collected, processed, or distributed. No human subjects, no scraped web data, no third-party datasets are involved. The dataset poses zero individual privacy risk by construction and contains no information that could be used to identify any natural person.",
   "rai:dataUseCases": "Intended use cases: (i) benchmarking training-policy selection methods (probe-then-commit routers, meta-routers, contextual bandits, learning-curve extrapolators) in hybrid PINN-operator pipelines; (ii) controlled PDE-level ablations of pretraining, causal loss, physics-weight balance, and gating-architecture choices; (iii) studying the accuracy-regret dissociation in evaluation-by-accuracy versus evaluation-by-regret. Out-of-scope: production PDE solving, safety-critical engineering deployment (climate forecasting, structural engineering, fluid simulation for aerospace), noisy-measurement or sparse-observation problems, and any clinical, financial, or surveillance application.",
@@ -11136,6 +11136,7 @@
   "rai:dataAnnotationProtocol": "No human-provided annotations. Each run is automatically labeled with its (PDE, configuration, policy, seed, stage) tuple at logging time. Statistical aggregations (mean, std, paired Wilcoxon p-value, Cohen's d, BH-FDR corrected significance) are computed by experiments/scripts/statistical_analysis.py.",
   "rai:safetyMeasures": "PDE solvers are abstract numerical methodology. Outputs are scalar error metrics on synthetic problems and pose no direct safety risk. Downstream users planning to apply hybrid PINN-operator solvers in safety-critical engineering (climate, fluid simulation, materials) should validate model fidelity and out-of-distribution robustness before deployment; the benchmark documents specific OOD-degradation findings (hybrid models degrade proportionally more than PINN-only on every evaluated coefficient/domain/IC shift) that practitioners should heed.",
   "rai:deidentificationMethod": "Not applicable. No identifying information present in source data.",
-  "rai:fairness": "The benchmark deliberately reports negative and conditional findings (causal loss helps with pretraining and hurts without it; routing accuracy and regret dissociate by 77x; diagnostic bound fails universally at probe window k=20 and on Burgers-type shock dynamics). Selection of PDEs spans diffusion, advection, nonlinear waves, dispersive, reaction-diffusion, and Navier-Stokes regimes to reduce family-specific bias. Family-level holdout (LOPO-F) is reported.",
-  "rai:releaseMaintenance": "Version 1.0.0 was frozen at submission (2026-05-04). Version 1.1.0 (camera-ready) adds the result sets reported in the camera-ready: a noise x sparsity sweep over six equation families, a d=3 Heat3D probe, a multi-scale MultiscaleHeat2D probe and five-policy suite, a non-PINN dissociation-transfer study, and the k=20 regret-bound check. Issues and pull requests: https://github.com/suanlab/PINNBench. 12-month issue/PR response commitment from camera-ready. Pinned PyTorch 2.5.1 and Python 3.12."
-}
+  "rai:fairness": "The benchmark deliberately reports negative and conditional findings (causal loss helps with pretraining and hurts without it; routing accuracy and regret dissociate by 77x; the diagnostic regret check fails on two Burgers configurations; its earlier reported failure at k=20 was a validator artifact). Selection of PDEs spans diffusion, advection, nonlinear waves, dispersive, reaction-diffusion, and Navier-Stokes regimes to reduce family-specific bias. A leave-one-family-out evaluation (each router retrained without the held-out family) is reported.",
+  "rai:releaseMaintenance": "Version 1.0.0 was frozen at submission (2026-05-04). Version 1.1.0 (camera-ready) adds the result sets reported in the camera-ready: a noise x sparsity sweep over six equation families, a d=3 Heat3D probe, a multi-scale MultiscaleHeat2D probe and five-policy suite, a non-PINN dissociation-transfer study, and the k=20 regret-bound check. Version 1.2.0 (2026-10-29) adds the leave-one-family-out routing runs (routing_family_holdout, meta_router_family_holdout), regenerated stage-ablation statistics over 9 PDEs with a corrected Benjamini-Hochberg step-up, and result sets cited by the paper that 1.1.0 omitted (hypino_matched_budget, pretrain_budget_sweep, clamping_ablation, noisy_data_probe, rd_speed sweeps, multiscale policy-suite parts). Issues and pull requests: https://github.com/suanlab/PINNBench. 12-month issue/PR response commitment from camera-ready. Pinned PyTorch 2.5.1 and Python 3.12.",
+  "dateModified": "2026-10-29"
+}

results/paper_c/ablation_statistics/results.json

@@ -18,8 +18,8 @@
     "wilcoxon_p": 0.02734375,
     "cohens_d": 0.9143632870287136,
     "interpretation": "Effect of adding causal loss when operator pretraining is present.",
-    "bh_fdr_threshold": 0.03541666666666667,
-    "bh_fdr_p_adjusted": 0.03860294117647059,
+    "bh_fdr_threshold": 0.037037037037037035,
+    "bh_fdr_p_adjusted": 0.0369140625,
     "bh_fdr_significant": true
   },
   {
@@ -41,8 +41,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": -2.1055616986457526,
     "interpretation": "Effect of adding causal loss when operator pretraining is skipped.",
-    "bh_fdr_threshold": 0.0020833333333333333,
-    "bh_fdr_p_adjusted": 0.046875,
+    "bh_fdr_threshold": 0.001851851851851852,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -64,8 +64,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": -23.132033999563973,
     "interpretation": "Effect of operator pretraining (stage 1) without causal loss.",
-    "bh_fdr_threshold": 0.004166666666666667,
-    "bh_fdr_p_adjusted": 0.0234375,
+    "bh_fdr_threshold": 0.003703703703703704,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -87,8 +87,8 @@
     "wilcoxon_p": 0.064453125,
     "cohens_d": 0.7725815820422364,
     "interpretation": "Effect of adding causal loss when operator pretraining is present.",
-    "bh_fdr_threshold": 0.03958333333333334,
-    "bh_fdr_p_adjusted": 0.08141447368421052,
+    "bh_fdr_threshold": 0.040740740740740744,
+    "bh_fdr_p_adjusted": 0.0791015625,
     "bh_fdr_significant": false
   },
   {
@@ -110,8 +110,8 @@
     "wilcoxon_p": 0.00390625,
     "cohens_d": 1.4198176815408456,
     "interpretation": "Effect of adding causal loss when operator pretraining is skipped.",
-    "bh_fdr_threshold": 0.02916666666666667,
-    "bh_fdr_p_adjusted": 0.006696428571428571,
+    "bh_fdr_threshold": 0.027777777777777776,
+    "bh_fdr_p_adjusted": 0.00703125,
     "bh_fdr_significant": true
   },
   {
@@ -133,8 +133,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": -4.940132204992188,
     "interpretation": "Effect of operator pretraining (stage 1) without causal loss.",
-    "bh_fdr_threshold": 0.006250000000000001,
-    "bh_fdr_p_adjusted": 0.015625,
+    "bh_fdr_threshold": 0.005555555555555557,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -156,8 +156,8 @@
     "wilcoxon_p": 0.037109375,
     "cohens_d": 0.7806599248045102,
     "interpretation": "Effect of adding causal loss when operator pretraining is present.",
-    "bh_fdr_threshold": 0.0375,
-    "bh_fdr_p_adjusted": 0.049479166666666664,
+    "bh_fdr_threshold": 0.03888888888888889,
+    "bh_fdr_p_adjusted": 0.04771205357142857,
     "bh_fdr_significant": true
   },
   {
@@ -179,8 +179,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": 1.4111086378179183,
     "interpretation": "Effect of adding causal loss when operator pretraining is skipped.",
-    "bh_fdr_threshold": 0.008333333333333333,
-    "bh_fdr_p_adjusted": 0.01171875,
+    "bh_fdr_threshold": 0.007407407407407408,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -202,8 +202,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": -4.681868848490021,
     "interpretation": "Effect of operator pretraining (stage 1) without causal loss.",
-    "bh_fdr_threshold": 0.010416666666666666,
-    "bh_fdr_p_adjusted": 0.009375,
+    "bh_fdr_threshold": 0.009259259259259259,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -225,8 +225,8 @@
     "wilcoxon_p": 0.16015625,
     "cohens_d": 0.4964139706693682,
     "interpretation": "Effect of adding causal loss when operator pretraining is present.",
-    "bh_fdr_threshold": 0.043750000000000004,
-    "bh_fdr_p_adjusted": 0.18303571428571427,
+    "bh_fdr_threshold": 0.04444444444444445,
+    "bh_fdr_p_adjusted": 0.17296875,
     "bh_fdr_significant": false
   },
   {
@@ -248,8 +248,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": -2.391120339225372,
     "interpretation": "Effect of adding causal loss when operator pretraining is skipped.",
-    "bh_fdr_threshold": 0.012500000000000002,
-    "bh_fdr_p_adjusted": 0.0078125,
+    "bh_fdr_threshold": 0.011111111111111113,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -271,8 +271,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": -2.7233353753020104,
     "interpretation": "Effect of operator pretraining (stage 1) without causal loss.",
-    "bh_fdr_threshold": 0.014583333333333335,
-    "bh_fdr_p_adjusted": 0.006696428571428571,
+    "bh_fdr_threshold": 0.012962962962962964,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -294,7 +294,7 @@
     "wilcoxon_p": 0.4921875,
     "cohens_d": 0.2585043012361478,
     "interpretation": "Effect of adding causal loss when operator pretraining is present.",
-    "bh_fdr_threshold": 0.05000000000000001,
+    "bh_fdr_threshold": 0.05,
     "bh_fdr_p_adjusted": 0.4921875,
     "bh_fdr_significant": false
   },
@@ -317,8 +317,8 @@
     "wilcoxon_p": 0.16015625,
     "cohens_d": 0.6413838759603556,
     "interpretation": "Effect of adding causal loss when operator pretraining is skipped.",
-    "bh_fdr_threshold": 0.04583333333333334,
-    "bh_fdr_p_adjusted": 0.1747159090909091,
+    "bh_fdr_threshold": 0.046296296296296294,
+    "bh_fdr_p_adjusted": 0.17296875,
     "bh_fdr_significant": false
   },
   {
@@ -340,8 +340,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": -9.723749957644019,
     "interpretation": "Effect of operator pretraining (stage 1) without causal loss.",
-    "bh_fdr_threshold": 0.016666666666666666,
-    "bh_fdr_p_adjusted": 0.005859375,
+    "bh_fdr_threshold": 0.014814814814814815,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -363,8 +363,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": 1.3771003367468224,
     "interpretation": "Effect of adding causal loss when operator pretraining is present.",
-    "bh_fdr_threshold": 0.01875,
-    "bh_fdr_p_adjusted": 0.005208333333333333,
+    "bh_fdr_threshold": 0.016666666666666666,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -386,8 +386,8 @@
     "wilcoxon_p": 0.083984375,
     "cohens_d": -0.6328842551929618,
     "interpretation": "Effect of adding causal loss when operator pretraining is skipped.",
-    "bh_fdr_threshold": 0.041666666666666664,
-    "bh_fdr_p_adjusted": 0.10078125,
+    "bh_fdr_threshold": 0.0425925925925926,
+    "bh_fdr_p_adjusted": 0.09859035326086957,
     "bh_fdr_significant": false
   },
   {
@@ -409,8 +409,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": -6.425487689783985,
     "interpretation": "Effect of operator pretraining (stage 1) without causal loss.",
-    "bh_fdr_threshold": 0.020833333333333332,
-    "bh_fdr_p_adjusted": 0.0046875,
+    "bh_fdr_threshold": 0.018518518518518517,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -432,8 +432,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": 1.8714990967532308,
     "interpretation": "Effect of adding causal loss when operator pretraining is present.",
-    "bh_fdr_threshold": 0.02291666666666667,
-    "bh_fdr_p_adjusted": 0.004261363636363636,
+    "bh_fdr_threshold": 0.020370370370370372,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -455,8 +455,8 @@
     "wilcoxon_p": 0.232421875,
     "cohens_d": -0.4090918744528993,
     "interpretation": "Effect of adding causal loss when operator pretraining is skipped.",
-    "bh_fdr_threshold": 0.04791666666666667,
-    "bh_fdr_p_adjusted": 0.2425271739130435,
+    "bh_fdr_threshold": 0.04814814814814815,
+    "bh_fdr_p_adjusted": 0.2413611778846154,
     "bh_fdr_significant": false
   },
   {
@@ -478,8 +478,77 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": -6.756701147577776,
     "interpretation": "Effect of operator pretraining (stage 1) without causal loss.",
-    "bh_fdr_threshold": 0.025000000000000005,
-    "bh_fdr_p_adjusted": 0.00390625,
+    "bh_fdr_threshold": 0.022222222222222227,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
+    "bh_fdr_significant": true
+  },
+  {
+    "pde": "navier_stokes2d",
+    "comparison": "Causal WITH pretrain",
+    "group_a": "full_pipeline_no_causal",
+    "group_b": "full_pipeline_with_causal",
+    "metric": "pde_residual_norm",
+    "n_pairs": 10,
+    "mean_a": 0.07094420380890369,
+    "mean_b": 0.06752287670969963,
+    "mean_diff": 0.0034213270992040634,
+    "bootstrap_95ci": [
+      0.0018365353997796775,
+      0.004762738943099976
+    ],
+    "wilcoxon_w": 3.0,
+    "wilcoxon_w_plus": 52.0,
+    "wilcoxon_p": 0.009765625,
+    "cohens_d": 1.3806984114800414,
+    "interpretation": "Effect of adding causal loss when operator pretraining is present.",
+    "bh_fdr_threshold": 0.031481481481481485,
+    "bh_fdr_p_adjusted": 0.0146484375,
+    "bh_fdr_significant": true
+  },
+  {
+    "pde": "navier_stokes2d",
+    "comparison": "Causal WITHOUT pretrain",
+    "group_a": "no_pretrain_no_causal",
+    "group_b": "no_pretrain_with_causal",
+    "metric": "pde_residual_norm",
+    "n_pairs": 10,
+    "mean_a": 0.34725913405418396,
+    "mean_b": 0.340691465139389,
+    "mean_diff": 0.006567668914794922,
+    "bootstrap_95ci": [
+      0.0026342227309942247,
+      0.010307728126645087
+    ],
+    "wilcoxon_w": 5.0,
+    "wilcoxon_w_plus": 50.0,
+    "wilcoxon_p": 0.01953125,
+    "cohens_d": 1.0134989555577605,
+    "interpretation": "Effect of adding causal loss when operator pretraining is skipped.",
+    "bh_fdr_threshold": 0.03518518518518519,
+    "bh_fdr_p_adjusted": 0.027754934210526317,
+    "bh_fdr_significant": true
+  },
+  {
+    "pde": "navier_stokes2d",
+    "comparison": "Pretrain effect",
+    "group_a": "full_pipeline_no_causal",
+    "group_b": "no_pretrain_no_causal",
+    "metric": "pde_residual_norm",
+    "n_pairs": 10,
+    "mean_a": 0.07094420380890369,
+    "mean_b": 0.34725913405418396,
+    "mean_diff": -0.27631493024528025,
+    "bootstrap_95ci": [
+      -0.29321924841962754,
+      -0.2575608641840518
+    ],
+    "wilcoxon_w": 0.0,
+    "wilcoxon_w_plus": 0.0,
+    "wilcoxon_p": 0.001953125,
+    "cohens_d": -9.037902741179419,
+    "interpretation": "Effect of operator pretraining (stage 1) without causal loss.",
+    "bh_fdr_threshold": 0.024074074074074074,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   },
   {
@@ -524,8 +593,8 @@
     "wilcoxon_p": 0.005859375,
     "cohens_d": -1.0165913027303293,
     "interpretation": "Effect of adding causal loss when operator pretraining is skipped.",
-    "bh_fdr_threshold": 0.03125,
-    "bh_fdr_p_adjusted": 0.009375,
+    "bh_fdr_threshold": 0.02962962962962963,
+    "bh_fdr_p_adjusted": 0.0098876953125,
     "bh_fdr_significant": true
   },
   {
@@ -547,8 +616,8 @@
     "wilcoxon_p": 0.001953125,
     "cohens_d": -3.3855038335316414,
     "interpretation": "Effect of operator pretraining (stage 1) without causal loss.",
-    "bh_fdr_threshold": 0.027083333333333334,
-    "bh_fdr_p_adjusted": 0.003605769230769231,
+    "bh_fdr_threshold": 0.02592592592592593,
+    "bh_fdr_p_adjusted": 0.0037667410714285715,
     "bh_fdr_significant": true
   }
 ]

results/paper_c/ablation_statistics/results.md

@@ -0,0 +1,63 @@
+# Paper C: Stage Ablation Statistical Analysis
+
+## Method
+
+- **Test**: Exact Wilcoxon signed-rank (paired by seed)
+- **CI**: Bootstrap 95% CI for paired mean difference (10,000 resamples)
+- **Effect size**: Paired Cohen's d
+- **Multiple-comparison correction**: Benjamini-Hochberg FDR (q=0.05) across all 24 tests
+- **Metric**: `pde_residual_norm` (lower is better)
+- **Convention**: diff = group_a - group_b (negative means A is better)
+
+## Results Table
+
+| PDE | Comparison | Mean diff | 95% CI | Wilcoxon p | BH p_adj | Cohen's d | Sig? |
+|---|---|---:|---:|---:|---:|---:|---|
+| advdiff1d | Causal WITH pretrain | 0.001568 | [0.000554, 0.002539] | 0.0273 | 0.0369 | 0.914 | yes |
+| advdiff1d | Causal WITHOUT pretrain | -0.024237 | [-0.030559, -0.017056] | 0.0020 | 0.0038 | -2.106 | yes |
+| advdiff1d | Pretrain effect | -0.164499 | [-0.168610, -0.160279] | 0.0020 | 0.0038 | -23.132 | yes |
+| allencahn1d | Causal WITH pretrain | 0.001753 | [0.000414, 0.003102] | 0.0645 | 0.0791 | 0.773 | no |
+| allencahn1d | Causal WITHOUT pretrain | 0.011799 | [0.007019, 0.016755] | 0.0039 | 0.0070 | 1.420 | yes |
+| allencahn1d | Pretrain effect | -0.187730 | [-0.212613, -0.168550] | 0.0020 | 0.0038 | -4.940 | yes |
+| allencahn1d_sharp | Causal WITH pretrain | 0.002366 | [0.000651, 0.004178] | 0.0371 | 0.0477 | 0.781 | yes |
+| allencahn1d_sharp | Causal WITHOUT pretrain | 0.010020 | [0.006050, 0.014341] | 0.0020 | 0.0038 | 1.411 | yes |
+| allencahn1d_sharp | Pretrain effect | -0.144906 | [-0.164763, -0.128635] | 0.0020 | 0.0038 | -4.682 | yes |
+| burgers1d | Causal WITH pretrain | 0.002397 | [-0.000520, 0.005047] | 0.1602 | 0.1730 | 0.496 | no |
+| burgers1d | Causal WITHOUT pretrain | -0.016872 | [-0.020800, -0.012489] | 0.0020 | 0.0038 | -2.391 | yes |
+| burgers1d | Pretrain effect | -0.099987 | [-0.119766, -0.076532] | 0.0020 | 0.0038 | -2.723 | yes |
+| burgers1d_verylow | Causal WITH pretrain | 0.000588 | [-0.000776, 0.001887] | 0.4922 | 0.4922 | 0.259 | no |
+| burgers1d_verylow | Causal WITHOUT pretrain | 0.002274 | [0.000322, 0.004439] | 0.1602 | 0.1730 | 0.641 | no |
+| burgers1d_verylow | Pretrain effect | -0.115856 | [-0.122849, -0.109043] | 0.0020 | 0.0038 | -9.724 | yes |
+| heat1d | Causal WITH pretrain | 0.001006 | [0.000630, 0.001483] | 0.0020 | 0.0038 | 1.377 | yes |
+| heat1d | Causal WITHOUT pretrain | -0.004095 | [-0.007723, -0.000142] | 0.0840 | 0.0986 | -0.633 | no |
+| heat1d | Pretrain effect | -0.089447 | [-0.098267, -0.082066] | 0.0020 | 0.0038 | -6.425 | yes |
+| kdv1d | Causal WITH pretrain | 0.001104 | [0.000774, 0.001461] | 0.0020 | 0.0038 | 1.871 | yes |
+| kdv1d | Causal WITHOUT pretrain | -0.002791 | [-0.006661, 0.001315] | 0.2324 | 0.2414 | -0.409 | no |
+| kdv1d | Pretrain effect | -0.127215 | [-0.138616, -0.116608] | 0.0020 | 0.0038 | -6.757 | yes |
+| navier_stokes2d | Causal WITH pretrain | 0.003421 | [0.001837, 0.004763] | 0.0098 | 0.0146 | 1.381 | yes |
+| navier_stokes2d | Causal WITHOUT pretrain | 0.006568 | [0.002634, 0.010308] | 0.0195 | 0.0278 | 1.013 | yes |
+| navier_stokes2d | Pretrain effect | -0.276315 | [-0.293219, -0.257561] | 0.0020 | 0.0038 | -9.038 | yes |
+| reaction_diffusion1d_stiff | Causal WITH pretrain | 0.003327 | [0.001539, 0.005298] | 0.0098 | 0.0146 | 1.028 | yes |
+| reaction_diffusion1d_stiff | Causal WITHOUT pretrain | -0.018012 | [-0.029249, -0.008687] | 0.0059 | 0.0099 | -1.017 | yes |
+| reaction_diffusion1d_stiff | Pretrain effect | -0.207585 | [-0.245946, -0.173969] | 0.0020 | 0.0038 | -3.386 | yes |
+
+## Summary
+
+- **Causal WITH pretrain**: 6/9 significant
+- **Causal WITHOUT pretrain**: 6/9 significant
+- **Pretrain effect**: 9/9 significant
+- **Overall**: 21/27 significant (BH-FDR q=0.05)
+
+## Interpretation Notes
+
+### Causal WITH pretrain
+- Significant: advdiff1d, allencahn1d_sharp, heat1d, kdv1d, navier_stokes2d, reaction_diffusion1d_stiff
+- Not significant: allencahn1d, burgers1d, burgers1d_verylow
+
+### Causal WITHOUT pretrain
+- Significant: advdiff1d, allencahn1d, allencahn1d_sharp, burgers1d, navier_stokes2d, reaction_diffusion1d_stiff
+- Not significant: burgers1d_verylow, heat1d, kdv1d
+
+### Pretrain effect
+- Significant: advdiff1d, allencahn1d, allencahn1d_sharp, burgers1d, burgers1d_verylow, heat1d, kdv1d, navier_stokes2d, reaction_diffusion1d_stiff
+

results/paper_combined/clamping_ablation/aggregate.json

@@ -0,0 +1,388 @@
+{
+  "experiment": "clamping_ablation",
+  "pdes": [
+    "heat1d",
+    "advdiff1d",
+    "burgers1d",
+    "burgers1d_verylow",
+    "allencahn1d",
+    "allencahn1d_sharp",
+    "kdv1d",
+    "reaction_diffusion1d",
+    "reaction_diffusion1d_stiff"
+  ],
+  "seeds": [
+    42,
+    123,
+    456,
+    789,
+    1024,
+    2048,
+    3072,
+    4096,
+    5120,
+    6144
+  ],
+  "by_pde": {
+    "heat1d": {
+      "unclamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 1.0571858406066894,
+          "std": 0.05756093107993466
+        },
+        "pde_residual_norm": {
+          "mean": 0.04358741119503975,
+          "std": 0.011486703180907037
+        },
+        "final_min_weight": {
+          "mean": 0.9971331357955933,
+          "std": 0.0014396344221535506
+        },
+        "training_time": {
+          "mean": 33.72910740375519,
+          "std": 0.8224657082370251
+        }
+      },
+      "clamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 1.0571858406066894,
+          "std": 0.05756093107993466
+        },
+        "pde_residual_norm": {
+          "mean": 0.04358741119503975,
+          "std": 0.011486703180907037
+        },
+        "final_min_weight": {
+          "mean": 0.9971331357955933,
+          "std": 0.0014396344221535506
+        },
+        "training_time": {
+          "mean": 33.44107687473297,
+          "std": 0.28333730725825207
+        }
+      }
+    },
+    "advdiff1d": {
+      "unclamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.9517185926437378,
+          "std": 0.020204631688324107
+        },
+        "pde_residual_norm": {
+          "mean": 0.060961409285664556,
+          "std": 0.00799904772115158
+        },
+        "final_min_weight": {
+          "mean": 0.9847445905208587,
+          "std": 0.005568130212210205
+        },
+        "training_time": {
+          "mean": 34.2068374633789,
+          "std": 0.5239867771253491
+        }
+      },
+      "clamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.9517185926437378,
+          "std": 0.020204631688324107
+        },
+        "pde_residual_norm": {
+          "mean": 0.060961409285664556,
+          "std": 0.00799904772115158
+        },
+        "final_min_weight": {
+          "mean": 0.9847445905208587,
+          "std": 0.005568130212210205
+        },
+        "training_time": {
+          "mean": 34.208179569244386,
+          "std": 0.42376675789271845
+        }
+      }
+    },
+    "burgers1d": {
+      "unclamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "pde_residual_norm": {
+          "mean": 0.12808200493454933,
+          "std": 0.005060598081042694
+        },
+        "final_min_weight": {
+          "mean": 0.9652827262878418,
+          "std": 0.008112499099729132
+        },
+        "training_time": {
+          "mean": 32.262592220306395,
+          "std": 1.082733815706432
+        }
+      },
+      "clamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "pde_residual_norm": {
+          "mean": 0.12808200493454933,
+          "std": 0.005060598081042694
+        },
+        "final_min_weight": {
+          "mean": 0.9652827262878418,
+          "std": 0.008112499099729132
+        },
+        "training_time": {
+          "mean": 32.57133233547211,
+          "std": 0.7608063207474912
+        }
+      }
+    },
+    "burgers1d_verylow": {
+      "unclamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "pde_residual_norm": {
+          "mean": 0.14588495790958406,
+          "std": 0.012864733495600947
+        },
+        "final_min_weight": {
+          "mean": 0.9768492460250855,
+          "std": 0.005148158181515563
+        },
+        "training_time": {
+          "mean": 31.973636269569397,
+          "std": 0.418574972011597
+        }
+      },
+      "clamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "pde_residual_norm": {
+          "mean": 0.14588495790958406,
+          "std": 0.012864733495600947
+        },
+        "final_min_weight": {
+          "mean": 0.9768492460250855,
+          "std": 0.005148158181515563
+        },
+        "training_time": {
+          "mean": 32.21248302459717,
+          "std": 0.9273376111782667
+        }
+      }
+    },
+    "allencahn1d": {
+      "unclamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.5255619168281556,
+          "std": 0.008502159073817284
+        },
+        "pde_residual_norm": {
+          "mean": 0.24164985865354538,
+          "std": 0.006022568669397903
+        },
+        "final_min_weight": {
+          "mean": 0.942301768064499,
+          "std": 0.003391796454175991
+        },
+        "training_time": {
+          "mean": 34.23547964096069,
+          "std": 0.9236276894410148
+        }
+      },
+      "clamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.5255619168281556,
+          "std": 0.008502159073817284
+        },
+        "pde_residual_norm": {
+          "mean": 0.24164985865354538,
+          "std": 0.006022568669397903
+        },
+        "final_min_weight": {
+          "mean": 0.942301768064499,
+          "std": 0.003391796454175991
+        },
+        "training_time": {
+          "mean": 33.579597449302675,
+          "std": 0.5929320395284455
+        }
+      }
+    },
+    "allencahn1d_sharp": {
+      "unclamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.533612996339798,
+          "std": 0.008274427321835108
+        },
+        "pde_residual_norm": {
+          "mean": 0.2339791998267174,
+          "std": 0.006644742445736239
+        },
+        "final_min_weight": {
+          "mean": 0.9454948306083679,
+          "std": 0.003433981008001039
+        },
+        "training_time": {
+          "mean": 34.0330765247345,
+          "std": 0.7840994811243875
+        }
+      },
+      "clamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.533612996339798,
+          "std": 0.008274427321835108
+        },
+        "pde_residual_norm": {
+          "mean": 0.2339791998267174,
+          "std": 0.006644742445736239
+        },
+        "final_min_weight": {
+          "mean": 0.9454948306083679,
+          "std": 0.003433981008001039
+        },
+        "training_time": {
+          "mean": 33.80888826847077,
+          "std": 1.0670600467543925
+        }
+      }
+    },
+    "kdv1d": {
+      "unclamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.42065784633159636,
+          "std": 0.04125193291983065
+        },
+        "pde_residual_norm": {
+          "mean": 0.11807976886630059,
+          "std": 0.012903259463317945
+        },
+        "final_min_weight": {
+          "mean": 0.9843865036964417,
+          "std": 0.0025105687284421472
+        },
+        "training_time": {
+          "mean": 34.82797448635101,
+          "std": 0.6712923082861436
+        }
+      },
+      "clamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.42065784633159636,
+          "std": 0.04125193291983065
+        },
+        "pde_residual_norm": {
+          "mean": 0.11807976886630059,
+          "std": 0.012903259463317945
+        },
+        "final_min_weight": {
+          "mean": 0.9843865036964417,
+          "std": 0.0025105687284421472
+        },
+        "training_time": {
+          "mean": 34.89519031047821,
+          "std": 0.5007306216660952
+        }
+      }
+    },
+    "reaction_diffusion1d": {
+      "unclamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "pde_residual_norm": {
+          "mean": 0.08361115008592605,
+          "std": 0.0018861010128000328
+        },
+        "final_min_weight": {
+          "mean": 0.9849636077880859,
+          "std": 0.0069608234319692855
+        },
+        "training_time": {
+          "mean": 32.31012737751007,
+          "std": 0.8781451095972249
+        }
+      },
+      "clamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "pde_residual_norm": {
+          "mean": 0.08361115008592605,
+          "std": 0.0018861010128000328
+        },
+        "final_min_weight": {
+          "mean": 0.9849636077880859,
+          "std": 0.0069608234319692855
+        },
+        "training_time": {
+          "mean": 32.85067577362061,
+          "std": 0.5390291657596392
+        }
+      }
+    },
+    "reaction_diffusion1d_stiff": {
+      "unclamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "pde_residual_norm": {
+          "mean": 0.21276859939098358,
+          "std": 0.010157276522028085
+        },
+        "final_min_weight": {
+          "mean": 0.9539549350738525,
+          "std": 0.0063930579867070465
+        },
+        "training_time": {
+          "mean": 32.403284549713135,
+          "std": 1.0721933135752917
+        }
+      },
+      "clamped": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "pde_residual_norm": {
+          "mean": 0.21276859939098358,
+          "std": 0.010157276522028085
+        },
+        "final_min_weight": {
+          "mean": 0.9539549350738525,
+          "std": 0.0063930579867070465
+        },
+        "training_time": {
+          "mean": 34.75156247615814,
+          "std": 0.56760320418834
+        }
+      }
+    }
+  }
+}

results/paper_combined/clamping_ablation/progress.json

@@ -0,0 +1,15 @@
+{
+  "completed_pdes": [
+    "heat1d",
+    "advdiff1d",
+    "burgers1d",
+    "burgers1d_verylow",
+    "allencahn1d",
+    "allencahn1d_sharp",
+    "kdv1d",
+    "reaction_diffusion1d",
+    "reaction_diffusion1d_stiff"
+  ],
+  "total_pdes": 9,
+  "last_completed": "reaction_diffusion1d_stiff"
+}

results/paper_combined/hypino_matched_budget/advdiff1d_1024_results.json

@@ -0,0 +1,50 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "advdiff1d",
+  "pde_config": {
+    "velocity": 1.0,
+    "alpha": 0.1,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 1024,
+  "matched_budget_seconds": 11.484827001113445,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 11.484827001113445,
+      "metrics": {
+        "relative_l2_error": 0.004154315063394714,
+        "mse": 1.1055341263217458e-06,
+        "mae": 0.0007217986143627501,
+        "max_error": 0.013241499662399297
+      },
+      "pde": "advdiff1d",
+      "seed": 1024,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 11.493239990901202,
+      "stage1_time": 2.917976802913472,
+      "stage1_steps": 200,
+      "stage2_time": 6.270509695168585,
+      "stage2_steps": 385,
+      "stage3_time": 2.3047534928191453,
+      "stage3_steps": 17,
+      "metrics": {
+        "pde_residual_norm": 0.12476573884487152,
+        "relative_l2_error": 0.4772602915763855,
+        "mse": 0.01368684321641922,
+        "mae": 0.07654865831136703,
+        "max_error": 0.6261355876922607,
+        "bc_violation": 0.031320180743932724,
+        "ic_violation": 0.13605216145515442
+      },
+      "pde": "advdiff1d",
+      "seed": 1024,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/advdiff1d_123_results.json

@@ -0,0 +1,50 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "advdiff1d",
+  "pde_config": {
+    "velocity": 1.0,
+    "alpha": 0.1,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 123,
+  "matched_budget_seconds": 11.165114372037351,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 11.165114372037351,
+      "metrics": {
+        "relative_l2_error": 0.00520255625870546,
+        "mse": 1.7338313963858167e-06,
+        "mae": 0.0009079280638130849,
+        "max_error": 0.012518852949142461
+      },
+      "pde": "advdiff1d",
+      "seed": 123,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 11.181147360010073,
+      "stage1_time": 3.5397472179029137,
+      "stage1_steps": 200,
+      "stage2_time": 5.395989370997995,
+      "stage2_steps": 348,
+      "stage3_time": 2.2454107711091638,
+      "stage3_steps": 16,
+      "metrics": {
+        "pde_residual_norm": 0.16972093284130096,
+        "relative_l2_error": 0.3457674980163574,
+        "mse": 0.007420359179377556,
+        "mae": 0.05547170341014862,
+        "max_error": 0.4359824061393738,
+        "bc_violation": 0.028230786323547363,
+        "ic_violation": 0.07177180051803589
+      },
+      "pde": "advdiff1d",
+      "seed": 123,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/advdiff1d_2048_results.json

@@ -0,0 +1,50 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "advdiff1d",
+  "pde_config": {
+    "velocity": 1.0,
+    "alpha": 0.1,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 2048,
+  "matched_budget_seconds": 11.460927323903888,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 11.460927323903888,
+      "metrics": {
+        "relative_l2_error": 0.004272311389345368,
+        "mse": 1.16922768577861e-06,
+        "mae": 0.0007068423212938393,
+        "max_error": 0.0135052502155304
+      },
+      "pde": "advdiff1d",
+      "seed": 2048,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 11.49724542023614,
+      "stage1_time": 3.004334368975833,
+      "stage1_steps": 200,
+      "stage2_time": 6.173016808228567,
+      "stage2_steps": 374,
+      "stage3_time": 2.31989424303174,
+      "stage3_steps": 17,
+      "metrics": {
+        "pde_residual_norm": 0.15798845887184143,
+        "relative_l2_error": 0.1659635454416275,
+        "mse": 0.0016460850602015853,
+        "mae": 0.0301732886582613,
+        "max_error": 0.1814490556716919,
+        "bc_violation": 0.0053941635414958,
+        "ic_violation": 0.01359289139509201
+      },
+      "pde": "advdiff1d",
+      "seed": 2048,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/advdiff1d_3072_results.json

@@ -0,0 +1,50 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "advdiff1d",
+  "pde_config": {
+    "velocity": 1.0,
+    "alpha": 0.1,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 3072,
+  "matched_budget_seconds": 10.919256535824388,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 10.919256535824388,
+      "metrics": {
+        "relative_l2_error": 0.004856509839956861,
+        "mse": 1.510851703234475e-06,
+        "mae": 0.0009005808143421949,
+        "max_error": 0.013477802276611333
+      },
+      "pde": "advdiff1d",
+      "seed": 3072,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 10.972043328918517,
+      "stage1_time": 3.156964933965355,
+      "stage1_steps": 200,
+      "stage2_time": 5.578621850116178,
+      "stage2_steps": 330,
+      "stage3_time": 2.236456544836983,
+      "stage3_steps": 17,
+      "metrics": {
+        "pde_residual_norm": 0.0811695009469986,
+        "relative_l2_error": 0.5975013375282288,
+        "mse": 0.0222932156175375,
+        "mae": 0.10211114585399628,
+        "max_error": 0.6729632616043091,
+        "bc_violation": 0.06646544486284256,
+        "ic_violation": 0.18444956839084625
+      },
+      "pde": "advdiff1d",
+      "seed": 3072,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/advdiff1d_4096_results.json

@@ -0,0 +1,50 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "advdiff1d",
+  "pde_config": {
+    "velocity": 1.0,
+    "alpha": 0.1,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 4096,
+  "matched_budget_seconds": 12.013140595983714,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 12.013140595983714,
+      "metrics": {
+        "relative_l2_error": 0.0041158834116111626,
+        "mse": 1.0851741027972953e-06,
+        "mae": 0.0007124984665854354,
+        "max_error": 0.013472944498062139
+      },
+      "pde": "advdiff1d",
+      "seed": 4096,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 12.085987030994147,
+      "stage1_time": 3.4215781409293413,
+      "stage1_steps": 200,
+      "stage2_time": 6.190741111058742,
+      "stage2_steps": 377,
+      "stage3_time": 2.473667779006064,
+      "stage3_steps": 18,
+      "metrics": {
+        "pde_residual_norm": 0.06396114081144333,
+        "relative_l2_error": 0.7088375687599182,
+        "mse": 0.027947237715125084,
+        "mae": 0.12363489717245102,
+        "max_error": 0.6397050619125366,
+        "bc_violation": 0.05193207040429115,
+        "ic_violation": 0.1991066336631775
+      },
+      "pde": "advdiff1d",
+      "seed": 4096,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/advdiff1d_42_results.json

@@ -0,0 +1,50 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "advdiff1d",
+  "pde_config": {
+    "velocity": 1.0,
+    "alpha": 0.1,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 42,
+  "matched_budget_seconds": 11.918144754134119,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 11.918144754134119,
+      "metrics": {
+        "relative_l2_error": 0.006033599794970775,
+        "mse": 2.3319879022815214e-06,
+        "mae": 0.0009231620735582799,
+        "max_error": 0.012683182954788213
+      },
+      "pde": "advdiff1d",
+      "seed": 42,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 11.92934147734195,
+      "stage1_time": 3.141978193074465,
+      "stage1_steps": 200,
+      "stage2_time": 6.401548201218247,
+      "stage2_steps": 384,
+      "stage3_time": 2.3858150830492377,
+      "stage3_steps": 17,
+      "metrics": {
+        "pde_residual_norm": 0.10928227007389069,
+        "relative_l2_error": 0.4801235795021057,
+        "mse": 0.013171485625207424,
+        "mae": 0.07329960912466049,
+        "max_error": 0.7680358290672302,
+        "bc_violation": 0.03402978181838989,
+        "ic_violation": 0.14542916417121887
+      },
+      "pde": "advdiff1d",
+      "seed": 42,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/advdiff1d_456_results.json

@@ -0,0 +1,50 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "advdiff1d",
+  "pde_config": {
+    "velocity": 1.0,
+    "alpha": 0.1,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 456,
+  "matched_budget_seconds": 11.614140221849084,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 11.614140221849084,
+      "metrics": {
+        "relative_l2_error": 0.0039440385712149785,
+        "mse": 9.96450202000395e-07,
+        "mae": 0.0006230241642957226,
+        "max_error": 0.013666480779647832
+      },
+      "pde": "advdiff1d",
+      "seed": 456,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 11.674075645161793,
+      "stage1_time": 3.14056647894904,
+      "stage1_steps": 200,
+      "stage2_time": 6.164740638108924,
+      "stage2_steps": 378,
+      "stage3_time": 2.3687685281038284,
+      "stage3_steps": 16,
+      "metrics": {
+        "pde_residual_norm": 0.17051230370998383,
+        "relative_l2_error": 0.3382963240146637,
+        "mse": 0.007502552587538958,
+        "mae": 0.05539253354072571,
+        "max_error": 0.5182910561561584,
+        "bc_violation": 0.020495519042015076,
+        "ic_violation": 0.06807474792003632
+      },
+      "pde": "advdiff1d",
+      "seed": 456,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/advdiff1d_5120_results.json

@@ -0,0 +1,50 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "advdiff1d",
+  "pde_config": {
+    "velocity": 1.0,
+    "alpha": 0.1,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 5120,
+  "matched_budget_seconds": 11.267948973923922,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 11.267948973923922,
+      "metrics": {
+        "relative_l2_error": 0.00617004625524269,
+        "mse": 2.4386536959356257e-06,
+        "mae": 0.0009542724964430616,
+        "max_error": 0.014151006937026978
+      },
+      "pde": "advdiff1d",
+      "seed": 5120,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 11.392662419937551,
+      "stage1_time": 3.2711690100841224,
+      "stage1_steps": 200,
+      "stage2_time": 5.7463704058900476,
+      "stage2_steps": 358,
+      "stage3_time": 2.375123003963381,
+      "stage3_steps": 18,
+      "metrics": {
+        "pde_residual_norm": 0.11485396325588226,
+        "relative_l2_error": 0.5430392622947693,
+        "mse": 0.019273757934570312,
+        "mae": 0.09487166255712509,
+        "max_error": 0.7661759853363037,
+        "bc_violation": 0.04286846145987511,
+        "ic_violation": 0.16916364431381226
+      },
+      "pde": "advdiff1d",
+      "seed": 5120,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/advdiff1d_6144_results.json

@@ -0,0 +1,50 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "advdiff1d",
+  "pde_config": {
+    "velocity": 1.0,
+    "alpha": 0.1,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 6144,
+  "matched_budget_seconds": 11.903233736986294,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 11.903233736986294,
+      "metrics": {
+        "relative_l2_error": 0.004183642964186974,
+        "mse": 1.1211985339336773e-06,
+        "mae": 0.0006767754901849923,
+        "max_error": 0.011527210474014287
+      },
+      "pde": "advdiff1d",
+      "seed": 6144,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 11.942326032323763,
+      "stage1_time": 3.435958066023886,
+      "stage1_steps": 200,
+      "stage2_time": 6.092163710156456,
+      "stage2_steps": 337,
+      "stage3_time": 2.414204256143421,
+      "stage3_steps": 17,
+      "metrics": {
+        "pde_residual_norm": 0.15884418785572052,
+        "relative_l2_error": 0.2672363519668579,
+        "mse": 0.004279059823602438,
+        "mae": 0.03655892238020897,
+        "max_error": 0.5822622179985046,
+        "bc_violation": 0.01018791738897562,
+        "ic_violation": 0.06582993268966675
+      },
+      "pde": "advdiff1d",
+      "seed": 6144,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/advdiff1d_789_results.json

@@ -0,0 +1,50 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "advdiff1d",
+  "pde_config": {
+    "velocity": 1.0,
+    "alpha": 0.1,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 789,
+  "matched_budget_seconds": 11.116228773025796,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 11.116228773025796,
+      "metrics": {
+        "relative_l2_error": 0.004489236370467626,
+        "mse": 1.2909762066025973e-06,
+        "mae": 0.0007994268285674785,
+        "max_error": 0.012997388839721685
+      },
+      "pde": "advdiff1d",
+      "seed": 789,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 11.132311964873224,
+      "stage1_time": 3.175016899826005,
+      "stage1_steps": 200,
+      "stage2_time": 5.72271842090413,
+      "stage2_steps": 346,
+      "stage3_time": 2.2345766441430897,
+      "stage3_steps": 16,
+      "metrics": {
+        "pde_residual_norm": 0.13157789409160614,
+        "relative_l2_error": 0.3525516390800476,
+        "mse": 0.007114553824067116,
+        "mae": 0.05335530266165733,
+        "max_error": 0.54401695728302,
+        "bc_violation": 0.023894891142845154,
+        "ic_violation": 0.09423264116048813
+      },
+      "pde": "advdiff1d",
+      "seed": 789,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/aggregate.json

@@ -0,0 +1,203 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pdes": [
+    "heat1d",
+    "advdiff1d",
+    "burgers1d",
+    "allencahn1d"
+  ],
+  "seeds": [
+    42,
+    123,
+    456,
+    789,
+    1024,
+    2048,
+    3072,
+    4096,
+    5120,
+    6144
+  ],
+  "by_pde": {
+    "heat1d": {
+      "hypino_finetune": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.010757430705665575,
+          "std": 0.0006403713013101963
+        },
+        "wall_time": {
+          "mean": 11.18706406604033,
+          "std": 0.29785130836951695
+        }
+      },
+      "pinn_only_finetune": {
+        "n": 0,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "wall_time": {
+          "mean": 0.0,
+          "std": 0.0
+        }
+      },
+      "hybrid_matched": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.21628479659557343,
+          "std": 0.050216972789960776
+        },
+        "wall_time": {
+          "mean": 11.286812606477179,
+          "std": 0.2959898714601055
+        }
+      }
+    },
+    "advdiff1d": {
+      "hypino_finetune": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.0047422139919096615,
+          "std": 0.0007683207456568901
+        },
+        "wall_time": {
+          "mean": 11.4862962288782,
+          "std": 0.35555619701840446
+        }
+      },
+      "pinn_only_finetune": {
+        "n": 0,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "wall_time": {
+          "mean": 0.0,
+          "std": 0.0
+        }
+      },
+      "hybrid_matched": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.4276577398180962,
+          "std": 0.15523729926981855
+        },
+        "wall_time": {
+          "mean": 11.530038067069835,
+          "std": 0.3566260063856941
+        }
+      }
+    },
+    "burgers1d": {
+      "hypino_finetune": {
+        "n": 0,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "wall_time": {
+          "mean": 0.0,
+          "std": 0.0
+        }
+      },
+      "pinn_only_finetune": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "wall_time": {
+          "mean": 7.082439509220421,
+          "std": 0.34526826024328844
+        }
+      },
+      "hybrid_matched": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "wall_time": {
+          "mean": 7.1438857487635685,
+          "std": 0.3566736433036886
+        }
+      }
+    },
+    "allencahn1d": {
+      "hypino_finetune": {
+        "n": 0,
+        "relative_l2_error": {
+          "mean": NaN,
+          "std": NaN
+        },
+        "wall_time": {
+          "mean": 0.0,
+          "std": 0.0
+        }
+      },
+      "pinn_only_finetune": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.16810581609606742,
+          "std": 0.025539958672288306
+        },
+        "wall_time": {
+          "mean": 8.36898730401881,
+          "std": 0.23675629668306705
+        }
+      },
+      "hybrid_matched": {
+        "n": 10,
+        "relative_l2_error": {
+          "mean": 0.35060557425022126,
+          "std": 0.028315139685508337
+        },
+        "wall_time": {
+          "mean": 8.44877382745035,
+          "std": 0.21665514049685228
+        }
+      }
+    }
+  },
+  "overall": {
+    "hypino_finetune": {
+      "n": 20,
+      "relative_l2_error": {
+        "mean": 0.007749822348787618,
+        "std": 0.003089644658247346
+      },
+      "wall_time": {
+        "mean": 11.336680147459266,
+        "std": 0.3604896356165725
+      }
+    },
+    "pinn_only_finetune": {
+      "n": 20,
+      "relative_l2_error": {
+        "mean": 0.16810581609606742,
+        "std": 0.025539958672288306
+      },
+      "wall_time": {
+        "mean": 7.725713406619616,
+        "std": 0.7081194566246597
+      }
+    },
+    "hybrid_matched": {
+      "n": 40,
+      "relative_l2_error": {
+        "mean": 0.3315160368879636,
+        "std": 0.12949644370215754
+      },
+      "wall_time": {
+        "mean": 9.602377562440234,
+        "std": 1.891899023742758
+      }
+    }
+  },
+  "notes": [
+    "HyPINO branch is run on PDEs with closed-form mappings to HyPINO's two-variable operator vocabulary (Heat1D, AdvDiff1D).",
+    "Burgers1D and AllenCahn1D fall back to a PINN-only finetune baseline as the reference branch (HyPINO has no clean encoding).",
+    "Matched budget is enforced via wall-clock time on the same GPU."
+  ]
+}

results/paper_combined/hypino_matched_budget/allencahn1d_1024_results.json

@@ -0,0 +1,53 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "allencahn1d",
+  "pde_config": {
+    "epsilon": 0.01,
+    "length": 2.0,
+    "t_max": 1.0,
+    "front_center": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 1024,
+  "matched_budget_seconds": 8.69020267110318,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 8.69020267110318,
+      "metrics": {
+        "pde_residual_norm": 0.14059776067733765,
+        "relative_l2_error": 0.19309240579605103,
+        "mse": 0.03680391237139702,
+        "mae": 0.09867047518491745,
+        "max_error": 0.7637790441513062,
+        "bc_violation": 0.000610032060649246,
+        "ic_violation": 0.0424308255314827
+      },
+      "pde": "allencahn1d",
+      "seed": 1024,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 8.712115371832624,
+      "stage1_time": 3.219786247005686,
+      "stage1_steps": 200,
+      "stage2_time": 3.7446930478326976,
+      "stage2_steps": 227,
+      "stage3_time": 1.7476360769942403,
+      "stage3_steps": 13,
+      "metrics": {
+        "pde_residual_norm": 0.21039903163909912,
+        "relative_l2_error": 0.3335523009300232,
+        "mse": 0.10996997356414795,
+        "mae": 0.21199795603752136,
+        "max_error": 0.9694104194641113,
+        "bc_violation": 0.005667810328304768,
+        "ic_violation": 0.1169038712978363
+      },
+      "pde": "allencahn1d",
+      "seed": 1024,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/allencahn1d_123_results.json

@@ -0,0 +1,53 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "allencahn1d",
+  "pde_config": {
+    "epsilon": 0.01,
+    "length": 2.0,
+    "t_max": 1.0,
+    "front_center": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 123,
+  "matched_budget_seconds": 8.275335442041978,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 8.275335442041978,
+      "metrics": {
+        "pde_residual_norm": 0.08994114398956299,
+        "relative_l2_error": 0.12484853714704514,
+        "mse": 0.015355301089584827,
+        "mae": 0.048092104494571686,
+        "max_error": 0.6476908922195435,
+        "bc_violation": 0.00046423773164860904,
+        "ic_violation": 0.008223969489336014
+      },
+      "pde": "allencahn1d",
+      "seed": 123,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 8.394129257649183,
+      "stage1_time": 3.324965497944504,
+      "stage1_steps": 172,
+      "stage2_time": 3.3053355258889496,
+      "stage2_steps": 184,
+      "stage3_time": 1.7638282338157296,
+      "stage3_steps": 13,
+      "metrics": {
+        "pde_residual_norm": 0.20577113330364227,
+        "relative_l2_error": 0.34891000390052795,
+        "mse": 0.12014519423246384,
+        "mae": 0.22671900689601898,
+        "max_error": 0.9238227009773254,
+        "bc_violation": 0.0028411292005330324,
+        "ic_violation": 0.12400327622890472
+      },
+      "pde": "allencahn1d",
+      "seed": 123,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/allencahn1d_2048_results.json

@@ -0,0 +1,53 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "allencahn1d",
+  "pde_config": {
+    "epsilon": 0.01,
+    "length": 2.0,
+    "t_max": 1.0,
+    "front_center": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 2048,
+  "matched_budget_seconds": 8.381790076149628,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 8.381790076149628,
+      "metrics": {
+        "pde_residual_norm": 0.131626158952713,
+        "relative_l2_error": 0.18797075748443604,
+        "mse": 0.03486473858356476,
+        "mae": 0.09214463084936142,
+        "max_error": 0.7846686840057373,
+        "bc_violation": 0.0008053772035054862,
+        "ic_violation": 0.051227033138275146
+      },
+      "pde": "allencahn1d",
+      "seed": 2048,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 8.456024406012148,
+      "stage1_time": 3.3558986890129745,
+      "stage1_steps": 174,
+      "stage2_time": 3.360106524080038,
+      "stage2_steps": 214,
+      "stage3_time": 1.740019192919135,
+      "stage3_steps": 13,
+      "metrics": {
+        "pde_residual_norm": 0.20428773760795593,
+        "relative_l2_error": 0.3319786489009857,
+        "mse": 0.10801974684000015,
+        "mae": 0.21042142808437347,
+        "max_error": 0.9396284818649292,
+        "bc_violation": 0.0052551222033798695,
+        "ic_violation": 0.11990612000226974
+      },
+      "pde": "allencahn1d",
+      "seed": 2048,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/allencahn1d_3072_results.json

@@ -0,0 +1,53 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "allencahn1d",
+  "pde_config": {
+    "epsilon": 0.01,
+    "length": 2.0,
+    "t_max": 1.0,
+    "front_center": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 3072,
+  "matched_budget_seconds": 8.376085401978344,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 8.376085401978344,
+      "metrics": {
+        "pde_residual_norm": 0.12081234157085419,
+        "relative_l2_error": 0.19190824031829834,
+        "mse": 0.03640221059322357,
+        "mae": 0.09166521579027176,
+        "max_error": 0.816250205039978,
+        "bc_violation": 0.0005476025980897248,
+        "ic_violation": 0.027293896302580833
+      },
+      "pde": "allencahn1d",
+      "seed": 3072,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 8.481348962988704,
+      "stage1_time": 3.1578137271571904,
+      "stage1_steps": 200,
+      "stage2_time": 3.553917452925816,
+      "stage2_steps": 231,
+      "stage3_time": 1.7696177829056978,
+      "stage3_steps": 13,
+      "metrics": {
+        "pde_residual_norm": 0.21788595616817474,
+        "relative_l2_error": 0.4033891260623932,
+        "mse": 0.16024735569953918,
+        "mae": 0.2822113335132599,
+        "max_error": 0.9731259346008301,
+        "bc_violation": 0.003071751445531845,
+        "ic_violation": 0.18956992030143738
+      },
+      "pde": "allencahn1d",
+      "seed": 3072,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/allencahn1d_4096_results.json

@@ -0,0 +1,53 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "allencahn1d",
+  "pde_config": {
+    "epsilon": 0.01,
+    "length": 2.0,
+    "t_max": 1.0,
+    "front_center": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 4096,
+  "matched_budget_seconds": 8.1862393070478,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 8.1862393070478,
+      "metrics": {
+        "pde_residual_norm": 0.12990589439868927,
+        "relative_l2_error": 0.19413337111473083,
+        "mse": 0.03716432303190231,
+        "mae": 0.09414295107126236,
+        "max_error": 0.8172110319137573,
+        "bc_violation": 0.0001907499972730875,
+        "ic_violation": 0.035255853086709976
+      },
+      "pde": "allencahn1d",
+      "seed": 4096,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 8.288710586028174,
+      "stage1_time": 3.2161105000413954,
+      "stage1_steps": 200,
+      "stage2_time": 3.3345002830028534,
+      "stage2_steps": 218,
+      "stage3_time": 1.7380998029839247,
+      "stage3_steps": 13,
+      "metrics": {
+        "pde_residual_norm": 0.21253176033496857,
+        "relative_l2_error": 0.3965703547000885,
+        "mse": 0.15491127967834473,
+        "mae": 0.2732950448989868,
+        "max_error": 0.9550119638442993,
+        "bc_violation": 0.0011812546290457249,
+        "ic_violation": 0.13363973796367645
+      },
+      "pde": "allencahn1d",
+      "seed": 4096,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/allencahn1d_42_results.json

@@ -0,0 +1,53 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "allencahn1d",
+  "pde_config": {
+    "epsilon": 0.01,
+    "length": 2.0,
+    "t_max": 1.0,
+    "front_center": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 42,
+  "matched_budget_seconds": 8.810380891896784,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 8.810380891896784,
+      "metrics": {
+        "pde_residual_norm": 0.1116866022348404,
+        "relative_l2_error": 0.16233879327774048,
+        "mse": 0.026018694043159485,
+        "mae": 0.07349684834480286,
+        "max_error": 0.7456440329551697,
+        "bc_violation": 0.00022698446991853416,
+        "ic_violation": 0.03899892792105675
+      },
+      "pde": "allencahn1d",
+      "seed": 42,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 8.826622009975836,
+      "stage1_time": 3.179024836048484,
+      "stage1_steps": 200,
+      "stage2_time": 3.880354367895052,
+      "stage2_steps": 237,
+      "stage3_time": 1.7672428060323,
+      "stage3_steps": 13,
+      "metrics": {
+        "pde_residual_norm": 0.2115984857082367,
+        "relative_l2_error": 0.35127440094947815,
+        "mse": 0.12159884721040726,
+        "mae": 0.22837470471858978,
+        "max_error": 0.9826579093933105,
+        "bc_violation": 0.007198600564152002,
+        "ic_violation": 0.11764465272426605
+      },
+      "pde": "allencahn1d",
+      "seed": 42,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/allencahn1d_456_results.json

@@ -0,0 +1,53 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "allencahn1d",
+  "pde_config": {
+    "epsilon": 0.01,
+    "length": 2.0,
+    "t_max": 1.0,
+    "front_center": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 456,
+  "matched_budget_seconds": 7.945534503087401,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 7.945534503087401,
+      "metrics": {
+        "pde_residual_norm": 0.08528584986925125,
+        "relative_l2_error": 0.1344357281923294,
+        "mse": 0.01778709515929222,
+        "mae": 0.04767435044050217,
+        "max_error": 0.7414506673812866,
+        "bc_violation": 0.00010172139445785433,
+        "ic_violation": 0.024142513051629066
+      },
+      "pde": "allencahn1d",
+      "seed": 456,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 8.055096366908401,
+      "stage1_time": 3.1848246019799262,
+      "stage1_steps": 200,
+      "stage2_time": 3.187875298084691,
+      "stage2_steps": 190,
+      "stage3_time": 1.6823964668437839,
+      "stage3_steps": 12,
+      "metrics": {
+        "pde_residual_norm": 0.20864301919937134,
+        "relative_l2_error": 0.3597801923751831,
+        "mse": 0.12770289182662964,
+        "mae": 0.23746664822101593,
+        "max_error": 0.9330372214317322,
+        "bc_violation": 0.0011527485912665725,
+        "ic_violation": 0.1332341730594635
+      },
+      "pde": "allencahn1d",
+      "seed": 456,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/allencahn1d_5120_results.json

@@ -0,0 +1,53 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "allencahn1d",
+  "pde_config": {
+    "epsilon": 0.01,
+    "length": 2.0,
+    "t_max": 1.0,
+    "front_center": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 5120,
+  "matched_budget_seconds": 8.194794670911506,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 8.194794670911506,
+      "metrics": {
+        "pde_residual_norm": 0.09293028712272644,
+        "relative_l2_error": 0.14673684537410736,
+        "mse": 0.02128794975578785,
+        "mae": 0.056395746767520905,
+        "max_error": 0.6887547969818115,
+        "bc_violation": 0.00033861270640045404,
+        "ic_violation": 0.01841440051794052
+      },
+      "pde": "allencahn1d",
+      "seed": 5120,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 8.212532497942448,
+      "stage1_time": 2.9855500429403037,
+      "stage1_steps": 200,
+      "stage2_time": 3.5836233790032566,
+      "stage2_steps": 220,
+      "stage3_time": 1.6433590759988874,
+      "stage3_steps": 12,
+      "metrics": {
+        "pde_residual_norm": 0.22312690317630768,
+        "relative_l2_error": 0.34605491161346436,
+        "mse": 0.1183275356888771,
+        "mae": 0.23045437037944794,
+        "max_error": 0.9110562801361084,
+        "bc_violation": 0.007878029718995094,
+        "ic_violation": 0.12659907341003418
+      },
+      "pde": "allencahn1d",
+      "seed": 5120,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/allencahn1d_6144_results.json

@@ -0,0 +1,53 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "allencahn1d",
+  "pde_config": {
+    "epsilon": 0.01,
+    "length": 2.0,
+    "t_max": 1.0,
+    "front_center": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 6144,
+  "matched_budget_seconds": 8.358624323969707,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 8.358624323969707,
+      "metrics": {
+        "pde_residual_norm": 0.10317305475473404,
+        "relative_l2_error": 0.1538073718547821,
+        "mse": 0.023371638730168343,
+        "mae": 0.07399199903011322,
+        "max_error": 0.7260705232620239,
+        "bc_violation": 0.0009533590637147427,
+        "ic_violation": 0.026320431381464005
+      },
+      "pde": "allencahn1d",
+      "seed": 6144,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 8.489468059968203,
+      "stage1_time": 3.3503546768333763,
+      "stage1_steps": 193,
+      "stage2_time": 3.3443585950881243,
+      "stage2_steps": 214,
+      "stage3_time": 1.7947547880467027,
+      "stage3_steps": 13,
+      "metrics": {
+        "pde_residual_norm": 0.194709911942482,
+        "relative_l2_error": 0.3259473443031311,
+        "mse": 0.10414231568574905,
+        "mae": 0.20026017725467682,
+        "max_error": 0.9560762047767639,
+        "bc_violation": 0.006463299039751291,
+        "ic_violation": 0.10573817789554596
+      },
+      "pde": "allencahn1d",
+      "seed": 6144,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/allencahn1d_789_results.json

@@ -0,0 +1,53 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "allencahn1d",
+  "pde_config": {
+    "epsilon": 0.01,
+    "length": 2.0,
+    "t_max": 1.0,
+    "front_center": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 789,
+  "matched_budget_seconds": 8.470885752001777,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 8.470885752001777,
+      "metrics": {
+        "pde_residual_norm": 0.1393750011920929,
+        "relative_l2_error": 0.19178611040115356,
+        "mse": 0.03632728010416031,
+        "mae": 0.09527671337127686,
+        "max_error": 0.8337691426277161,
+        "bc_violation": 0.0008994344971142709,
+        "ic_violation": 0.04890268296003342
+      },
+      "pde": "allencahn1d",
+      "seed": 789,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 8.571690755197778,
+      "stage1_time": 3.395089189056307,
+      "stage1_steps": 193,
+      "stage2_time": 3.3906155191361904,
+      "stage2_steps": 207,
+      "stage3_time": 1.7859860470052809,
+      "stage3_steps": 13,
+      "metrics": {
+        "pde_residual_norm": 0.2075824737548828,
+        "relative_l2_error": 0.30859845876693726,
+        "mse": 0.09387107938528061,
+        "mae": 0.19261391460895538,
+        "max_error": 0.9470365643501282,
+        "bc_violation": 0.005797241348773241,
+        "ic_violation": 0.111391581594944
+      },
+      "pde": "allencahn1d",
+      "seed": 789,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/burgers1d_1024_results.json

@@ -0,0 +1,44 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "burgers1d",
+  "pde_config": {
+    "viscosity": 0.01,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 1024,
+  "matched_budget_seconds": 6.616038148058578,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 6.616038148058578,
+      "metrics": {
+        "pde_residual_norm": 0.20441006124019623,
+        "bc_violation": 0.017613792791962624,
+        "ic_violation": 0.038427386432886124
+      },
+      "pde": "burgers1d",
+      "seed": 1024,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 6.679805057123303,
+      "stage1_time": 2.656797207193449,
+      "stage1_steps": 180,
+      "stage2_time": 2.644893816905096,
+      "stage2_steps": 167,
+      "stage3_time": 1.378114033024758,
+      "stage3_steps": 10,
+      "metrics": {
+        "pde_residual_norm": 0.22550320625305176,
+        "bc_violation": 0.011805089190602303,
+        "ic_violation": 0.05072234943509102
+      },
+      "pde": "burgers1d",
+      "seed": 1024,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/burgers1d_123_results.json

@@ -0,0 +1,44 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "burgers1d",
+  "pde_config": {
+    "viscosity": 0.01,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 123,
+  "matched_budget_seconds": 7.079934409121051,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 7.079934409121051,
+      "metrics": {
+        "pde_residual_norm": 0.19153529405593872,
+        "bc_violation": 0.013764698058366776,
+        "ic_violation": 0.04640496149659157
+      },
+      "pde": "burgers1d",
+      "seed": 123,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 7.15582737699151,
+      "stage1_time": 2.653643890051171,
+      "stage1_steps": 200,
+      "stage2_time": 3.016962236026302,
+      "stage2_steps": 227,
+      "stage3_time": 1.4852212509140372,
+      "stage3_steps": 11,
+      "metrics": {
+        "pde_residual_norm": 0.1938847452402115,
+        "bc_violation": 0.015817711129784584,
+        "ic_violation": 0.04979739710688591
+      },
+      "pde": "burgers1d",
+      "seed": 123,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/burgers1d_2048_results.json

@@ -0,0 +1,44 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "burgers1d",
+  "pde_config": {
+    "viscosity": 0.01,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 2048,
+  "matched_budget_seconds": 6.9327059329953045,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 6.9327059329953045,
+      "metrics": {
+        "pde_residual_norm": 0.1712341159582138,
+        "bc_violation": 0.008218296803534031,
+        "ic_violation": 0.04494340345263481
+      },
+      "pde": "burgers1d",
+      "seed": 2048,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 6.938157436205074,
+      "stage1_time": 2.7295828310307115,
+      "stage1_steps": 200,
+      "stage2_time": 2.8208878689911216,
+      "stage2_steps": 193,
+      "stage3_time": 1.387686736183241,
+      "stage3_steps": 10,
+      "metrics": {
+        "pde_residual_norm": 0.20011095702648163,
+        "bc_violation": 0.016325926408171654,
+        "ic_violation": 0.04804155230522156
+      },
+      "pde": "burgers1d",
+      "seed": 2048,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/burgers1d_3072_results.json

@@ -0,0 +1,44 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "burgers1d",
+  "pde_config": {
+    "viscosity": 0.01,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 3072,
+  "matched_budget_seconds": 7.149956083158031,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 7.149956083158031,
+      "metrics": {
+        "pde_residual_norm": 0.20724347233772278,
+        "bc_violation": 0.01770992949604988,
+        "ic_violation": 0.03796419873833656
+      },
+      "pde": "burgers1d",
+      "seed": 3072,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 7.1924566039815545,
+      "stage1_time": 2.8692238240037113,
+      "stage1_steps": 200,
+      "stage2_time": 2.855879121925682,
+      "stage2_steps": 196,
+      "stage3_time": 1.4673536580521613,
+      "stage3_steps": 10,
+      "metrics": {
+        "pde_residual_norm": 0.22016996145248413,
+        "bc_violation": 0.014933464117348194,
+        "ic_violation": 0.05736216530203819
+      },
+      "pde": "burgers1d",
+      "seed": 3072,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/burgers1d_4096_results.json

@@ -0,0 +1,44 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "burgers1d",
+  "pde_config": {
+    "viscosity": 0.01,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 4096,
+  "matched_budget_seconds": 7.495354883139953,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 7.495354883139953,
+      "metrics": {
+        "pde_residual_norm": 0.17734254896640778,
+        "bc_violation": 0.012503622099757195,
+        "ic_violation": 0.05469898506999016
+      },
+      "pde": "burgers1d",
+      "seed": 4096,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 7.58905250323005,
+      "stage1_time": 2.787865397054702,
+      "stage1_steps": 200,
+      "stage2_time": 3.2181000711862,
+      "stage2_steps": 217,
+      "stage3_time": 1.5830870349891484,
+      "stage3_steps": 12,
+      "metrics": {
+        "pde_residual_norm": 0.2090531587600708,
+        "bc_violation": 0.01590825617313385,
+        "ic_violation": 0.04608733579516411
+      },
+      "pde": "burgers1d",
+      "seed": 4096,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/burgers1d_42_results.json

@@ -0,0 +1,44 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "burgers1d",
+  "pde_config": {
+    "viscosity": 0.01,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 42,
+  "matched_budget_seconds": 7.116403200896457,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 7.116403200896457,
+      "metrics": {
+        "pde_residual_norm": 0.18602922558784485,
+        "bc_violation": 0.01844218000769615,
+        "ic_violation": 0.056280117481946945
+      },
+      "pde": "burgers1d",
+      "seed": 42,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 7.231115171220154,
+      "stage1_time": 2.8483544851187617,
+      "stage1_steps": 195,
+      "stage2_time": 2.853560324991122,
+      "stage2_steps": 212,
+      "stage3_time": 1.52920036111027,
+      "stage3_steps": 11,
+      "metrics": {
+        "pde_residual_norm": 0.20640622079372406,
+        "bc_violation": 0.01788671687245369,
+        "ic_violation": 0.047693345695734024
+      },
+      "pde": "burgers1d",
+      "seed": 42,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/burgers1d_456_results.json

@@ -0,0 +1,44 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "burgers1d",
+  "pde_config": {
+    "viscosity": 0.01,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 456,
+  "matched_budget_seconds": 6.672353338915855,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 6.672353338915855,
+      "metrics": {
+        "pde_residual_norm": 0.2008197009563446,
+        "bc_violation": 0.01739061065018177,
+        "ic_violation": 0.04686290770769119
+      },
+      "pde": "burgers1d",
+      "seed": 456,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 6.719709334196523,
+      "stage1_time": 2.642758168047294,
+      "stage1_steps": 200,
+      "stage2_time": 2.6956847989931703,
+      "stage2_steps": 194,
+      "stage3_time": 1.3812663671560585,
+      "stage3_steps": 10,
+      "metrics": {
+        "pde_residual_norm": 0.21323512494564056,
+        "bc_violation": 0.013567410409450531,
+        "ic_violation": 0.05053797736763954
+      },
+      "pde": "burgers1d",
+      "seed": 456,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/burgers1d_5120_results.json

@@ -0,0 +1,44 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "burgers1d",
+  "pde_config": {
+    "viscosity": 0.01,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 5120,
+  "matched_budget_seconds": 7.585222116904333,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 7.585222116904333,
+      "metrics": {
+        "pde_residual_norm": 0.18993298709392548,
+        "bc_violation": 0.02023250050842762,
+        "ic_violation": 0.0455697700381279
+      },
+      "pde": "burgers1d",
+      "seed": 5120,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 7.639178397133946,
+      "stage1_time": 2.7431522810366005,
+      "stage1_steps": 200,
+      "stage2_time": 3.3343359930440784,
+      "stage2_steps": 274,
+      "stage3_time": 1.5616901230532676,
+      "stage3_steps": 12,
+      "metrics": {
+        "pde_residual_norm": 0.20460620522499084,
+        "bc_violation": 0.016156604513525963,
+        "ic_violation": 0.05239883437752724
+      },
+      "pde": "burgers1d",
+      "seed": 5120,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/burgers1d_6144_results.json

@@ -0,0 +1,44 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "burgers1d",
+  "pde_config": {
+    "viscosity": 0.01,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 6144,
+  "matched_budget_seconds": 7.509465340990573,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 7.509465340990573,
+      "metrics": {
+        "pde_residual_norm": 0.19954417645931244,
+        "bc_violation": 0.018706869333982468,
+        "ic_violation": 0.04394349083304405
+      },
+      "pde": "burgers1d",
+      "seed": 6144,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 7.583197637926787,
+      "stage1_time": 3.010416015982628,
+      "stage1_steps": 197,
+      "stage2_time": 3.0061726649291813,
+      "stage2_steps": 198,
+      "stage3_time": 1.566608957014978,
+      "stage3_steps": 11,
+      "metrics": {
+        "pde_residual_norm": 0.21172556281089783,
+        "bc_violation": 0.013471605256199837,
+        "ic_violation": 0.05434518679976463
+      },
+      "pde": "burgers1d",
+      "seed": 6144,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/burgers1d_789_results.json

@@ -0,0 +1,44 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "burgers1d",
+  "pde_config": {
+    "viscosity": 0.01,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": false,
+  "seed": 789,
+  "matched_budget_seconds": 6.666961638024077,
+  "rows": [
+    {
+      "branch": "pinn_only_finetune",
+      "wall_time": 6.666961638024077,
+      "metrics": {
+        "pde_residual_norm": 0.19287538528442383,
+        "bc_violation": 0.024335922673344612,
+        "ic_violation": 0.041686687618494034
+      },
+      "pde": "burgers1d",
+      "seed": 789,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 6.710357969626784,
+      "stage1_time": 2.5617637138348073,
+      "stage1_steps": 200,
+      "stage2_time": 2.7801038469187915,
+      "stage2_steps": 201,
+      "stage3_time": 1.3684904088731855,
+      "stage3_steps": 10,
+      "metrics": {
+        "pde_residual_norm": 0.2093043476343155,
+        "bc_violation": 0.013944434002041817,
+        "ic_violation": 0.05870470777153969
+      },
+      "pde": "burgers1d",
+      "seed": 789,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/heat1d_1024_results.json

@@ -0,0 +1,49 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "heat1d",
+  "pde_config": {
+    "alpha": 1.0,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 1024,
+  "matched_budget_seconds": 10.872612959938124,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 10.872612959938124,
+      "metrics": {
+        "relative_l2_error": 0.011570004048959909,
+        "mse": 3.5115620144764196e-06,
+        "mae": 0.0015231783569025622,
+        "max_error": 0.009434163570404053
+      },
+      "pde": "heat1d",
+      "seed": 1024,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 11.02023670496419,
+      "stage1_time": 2.9072158720809966,
+      "stage1_steps": 200,
+      "stage2_time": 5.791842493927106,
+      "stage2_steps": 375,
+      "stage3_time": 2.321178338956088,
+      "stage3_steps": 17,
+      "metrics": {
+        "pde_residual_norm": 0.045260339975357056,
+        "relative_l2_error": 0.25442400574684143,
+        "mse": 0.0018389757024124265,
+        "mae": 0.037136584520339966,
+        "max_error": 0.15761689841747284,
+        "bc_violation": 0.0031206163112074137,
+        "ic_violation": 0.009072537533938885
+      },
+      "pde": "heat1d",
+      "seed": 1024,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/heat1d_123_results.json

@@ -0,0 +1,49 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "heat1d",
+  "pde_config": {
+    "alpha": 1.0,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 123,
+  "matched_budget_seconds": 11.509691530140117,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 11.509691530140117,
+      "metrics": {
+        "relative_l2_error": 0.010724938565217553,
+        "mse": 3.017330945599561e-06,
+        "mae": 0.0014064404640968355,
+        "max_error": 0.009019613265991305
+      },
+      "pde": "heat1d",
+      "seed": 123,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 11.651423167204484,
+      "stage1_time": 3.031029877020046,
+      "stage1_steps": 200,
+      "stage2_time": 6.193896782118827,
+      "stage2_steps": 376,
+      "stage3_time": 2.426496508065611,
+      "stage3_steps": 17,
+      "metrics": {
+        "pde_residual_norm": 0.056063853204250336,
+        "relative_l2_error": 0.15338872373104095,
+        "mse": 0.0006122124614194036,
+        "mae": 0.019447915256023407,
+        "max_error": 0.10658235847949982,
+        "bc_violation": 0.001646102755330503,
+        "ic_violation": 0.004867592826485634
+      },
+      "pde": "heat1d",
+      "seed": 123,
+      "success": true
+    }
+  ]
+}

results/paper_combined/hypino_matched_budget/heat1d_2048_results.json

@@ -0,0 +1,49 @@
+{
+  "experiment": "hypino_matched_budget",
+  "pde": "heat1d",
+  "pde_config": {
+    "alpha": 1.0,
+    "length": 1.0,
+    "t_max": 1.0
+  },
+  "hypino_supported": true,
+  "seed": 2048,
+  "matched_budget_seconds": 10.989870873978361,
+  "rows": [
+    {
+      "branch": "hypino_finetune",
+      "wall_time": 10.989870873978361,
+      "metrics": {
+        "relative_l2_error": 0.010144351370084152,
+        "mse": 2.6994910227953834e-06,
+        "mae": 0.001352712286409737,
+        "max_error": 0.008736908435821533
+      },
+      "pde": "heat1d",
+      "seed": 2048,
+      "success": true
+    },
+    {
+      "branch": "hybrid_matched",
+      "wall_time": 11.099930197000504,
+      "stage1_time": 3.126870078034699,
+      "stage1_steps": 200,
+      "stage2_time": 5.677308599930257,
+      "stage2_steps": 321,
+      "stage3_time": 2.295751519035548,
+      "stage3_steps": 17,
+      "metrics": {
+        "pde_residual_norm": 0.0532817468047142,
+        "relative_l2_error": 0.22200554609298706,
+        "mse": 0.0011161108268424869,
+        "mae": 0.02849474549293518,
+        "max_error": 0.13661596179008484,
+        "bc_violation": 0.0021929179783910513,
+        "ic_violation": 0.007334013935178518
+      },
+      "pde": "heat1d",
+      "seed": 2048,
+      "success": true
+    }
+  ]
+}