Add canonical experiment evidence

evidence/experiment_ledger.json

@@ -0,0 +1,274 @@
+{
+  "title": "IBM Quantum 150Q experiment ledger",
+  "canonical_repo_url": "https://github.com/sankalpsthakur/ibm-quantum-150q-ising",
+  "kaggle_url": "{{KAGGLE_URL}}",
+  "experiments": [
+    {
+      "id": "E00",
+      "execution": "classical + statevector",
+      "input": "Eight industrial assets; choose a subset that curtails exactly 12 MW with minimum disruption.",
+      "inference": "Exhaust all 256 schedules; separately optimize one-layer QAOA on a seeded statevector sampler.",
+      "outcome": "Exact score 10.4; QAOA recovered the same bitstring; exact-state probability 0.0034.",
+      "learning": "An auditable baseline is essential; this small instance proves correctness, not acceleration.",
+      "job_id": "\u2014",
+      "quantum_seconds": 0,
+      "status": "COMPLETED",
+      "evidence": "outputs/results.json"
+    },
+    {
+      "id": "E01",
+      "execution": "IBM hardware",
+      "input": "150-qubit GHZ preparation measured in Z and X bases; 1,024 shots per basis.",
+      "inference": "Estimate ideal all-equal Z population and global X parity as a two-basis coherence diagnostic.",
+      "outcome": "Ideal Z population 0.000977; X parity +0.046875.",
+      "learning": "Allocating 150 qubits did not preserve a high-fidelity cat state; this is not entanglement certification.",
+      "job_id": "d9gjobkhonhs73abnl3g",
+      "quantum_seconds": 3,
+      "status": "COMPLETED",
+      "evidence": "outputs/hardware_results.json"
+    },
+    {
+      "id": "E02",
+      "execution": "IBM hardware",
+      "input": "The same eight-asset, exactly-12-MW demand-response QUBO at the optimized statevector parameters.",
+      "inference": "Sample the compiled one-layer QAOA circuit and score every schedule classically.",
+      "outcome": "Feasible probability 0.1465; exact optimum sampled 0.0049 (5/1024).",
+      "learning": "Hardware can return the correct schedule, but the exact classical search remains faster for eight variables.",
+      "job_id": "d9gjobkhonhs73abnl3g",
+      "quantum_seconds": 0,
+      "status": "COMPLETED",
+      "evidence": "outputs/hardware_results.json"
+    },
+    {
+      "id": "E03",
+      "execution": "local simulator",
+      "input": "Compiled order finding for N=15 with base 2 on eight simulated qubits.",
+      "inference": "Use phase estimation to infer period 4, then classical GCD post-processing.",
+      "outcome": "Recovered factors 3 and 5 in 0.129s.",
+      "learning": "This validates the educational workflow only; it is not general or RSA-scale factorization.",
+      "job_id": "\u2014",
+      "quantum_seconds": 0,
+      "status": "COMPLETED",
+      "evidence": "outputs/hardware_results.json"
+    },
+    {
+      "id": "E04",
+      "execution": "classical exact",
+      "input": "150 binary variables, 167 weighted XOR constraints, total weight 873, deterministic seed 20260723.",
+      "inference": "Linearize XORs and solve the 317-variable MILP with SciPy/HiGHS at zero requested MIP gap.",
+      "outcome": "Exact 865/873 (0.9908); 161/167 constraints; MIP gap 0.0; 3.287s.",
+      "learning": "This is the production-quality answer and the truth baseline for every QPU sample.",
+      "job_id": "\u2014",
+      "quantum_seconds": 0,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_benchmark.json"
+    },
+    {
+      "id": "E05",
+      "execution": "IBM hardware",
+      "input": "One-layer 150Q QAOA; 100 beta/gamma points; 1,024 shots each; zero SWAPs.",
+      "inference": "Score 102,400 sampled 150-bit assignments and rank parameters by mean weighted objective.",
+      "outcome": "Best mean ratio 0.6733; best sample ratio 0.8035.",
+      "learning": "A broad hardware parameter sweep finds a useful region, but remains far below the exact classical optimum.",
+      "job_id": "d9gk5gchonhs73abo5ag",
+      "quantum_seconds": 30,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_benchmark.json"
+    },
+    {
+      "id": "E06",
+      "execution": "IBM hardware",
+      "input": "Ten leading p=1 points with gate/measurement twirling and XpXm dynamical decoupling; 4,096 shots each.",
+      "inference": "Re-score mitigated samples against the same exact optimum.",
+      "outcome": "Best mitigated mean ratio 0.5563; best sample 0.7110.",
+      "learning": "The combined mitigation stack degraded this circuit and required a controlled ablation.",
+      "job_id": "d9gk64ogk0ls73f1mar0",
+      "quantum_seconds": 14,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_benchmark.json"
+    },
+    {
+      "id": "E07",
+      "execution": "IBM hardware",
+      "input": "p=2 QAOA; 50 transferred/jittered parameter schedules; 4,096 shots each; 668 two-qubit gates; zero SWAPs.",
+      "inference": "Measure the mean and best-sample approximation ratios at a controlled circuit depth.",
+      "outcome": "Best mean ratio 0.6903; best sample ratio 0.8208; usage 57s.",
+      "learning": "Depth improves mean quality through p=6, then saturates/declines; isolated best samples are noisier than replicated means.",
+      "job_id": "d9gk9kjsbqfc73eoops0",
+      "quantum_seconds": 57,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_depth_and_mitigation.json"
+    },
+    {
+      "id": "E08",
+      "execution": "IBM hardware",
+      "input": "p=3 QAOA; 50 transferred/jittered parameter schedules; 4,096 shots each; 1002 two-qubit gates; zero SWAPs.",
+      "inference": "Measure the mean and best-sample approximation ratios at a controlled circuit depth.",
+      "outcome": "Best mean ratio 0.7048; best sample ratio 0.8439; usage 57s.",
+      "learning": "Depth improves mean quality through p=6, then saturates/declines; isolated best samples are noisier than replicated means.",
+      "job_id": "d9gkajl0k0jc738gsv80",
+      "quantum_seconds": 57,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_depth_and_mitigation.json"
+    },
+    {
+      "id": "E09",
+      "execution": "IBM hardware",
+      "input": "p=4 QAOA; 50 transferred/jittered parameter schedules; 4,096 shots each; 1336 two-qubit gates; zero SWAPs.",
+      "inference": "Measure the mean and best-sample approximation ratios at a controlled circuit depth.",
+      "outcome": "Best mean ratio 0.7263; best sample ratio 0.8578; usage 57s.",
+      "learning": "Depth improves mean quality through p=6, then saturates/declines; isolated best samples are noisier than replicated means.",
+      "job_id": "d9gkj98gk0ls73f1mspg",
+      "quantum_seconds": 57,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_extended_depth.json"
+    },
+    {
+      "id": "E10",
+      "execution": "IBM hardware",
+      "input": "p=5 QAOA; 50 transferred/jittered parameter schedules; 4,096 shots each; 1670 two-qubit gates; zero SWAPs.",
+      "inference": "Measure the mean and best-sample approximation ratios at a controlled circuit depth.",
+      "outcome": "Best mean ratio 0.7357; best sample ratio 0.8925; usage 58s.",
+      "learning": "Depth improves mean quality through p=6, then saturates/declines; isolated best samples are noisier than replicated means.",
+      "job_id": "d9gkle3sbqfc73eop9tg",
+      "quantum_seconds": 58,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_extended_depth.json"
+    },
+    {
+      "id": "E11",
+      "execution": "IBM hardware",
+      "input": "p=6 QAOA; 50 transferred/jittered parameter schedules; 4,096 shots each; 2004 two-qubit gates; zero SWAPs.",
+      "inference": "Measure the mean and best-sample approximation ratios at a controlled circuit depth.",
+      "outcome": "Best mean ratio 0.7390; best sample ratio 0.8775; usage 58s.",
+      "learning": "Depth improves mean quality through p=6, then saturates/declines; isolated best samples are noisier than replicated means.",
+      "job_id": "d9gknv4honhs73abouig",
+      "quantum_seconds": 58,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_extended_depth.json"
+    },
+    {
+      "id": "E12",
+      "execution": "IBM hardware",
+      "input": "p=7 QAOA; 50 transferred/jittered parameter schedules; 4,096 shots each; 2338 two-qubit gates; zero SWAPs.",
+      "inference": "Measure the mean and best-sample approximation ratios at a controlled circuit depth.",
+      "outcome": "Best mean ratio 0.7362; best sample ratio 0.8647; usage 58s.",
+      "learning": "Depth improves mean quality through p=6, then saturates/declines; isolated best samples are noisier than replicated means.",
+      "job_id": "d9gkq3ogk0ls73f1n5tg",
+      "quantum_seconds": 58,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_final_depth.json"
+    },
+    {
+      "id": "E13",
+      "execution": "IBM hardware",
+      "input": "p=8 QAOA; 50 transferred/jittered parameter schedules; 4,096 shots each; 2672 two-qubit gates; zero SWAPs.",
+      "inference": "Measure the mean and best-sample approximation ratios at a controlled circuit depth.",
+      "outcome": "Best mean ratio 0.7354; best sample ratio 0.8798; usage 58s.",
+      "learning": "Depth improves mean quality through p=6, then saturates/declines; isolated best samples are noisier than replicated means.",
+      "job_id": "d9gkrnjsbqfc73eopi3g",
+      "quantum_seconds": 58,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_final_depth.json"
+    },
+    {
+      "id": "E14",
+      "execution": "IBM hardware",
+      "input": "Fixed best p=1 parameters; No mitigation; 10 replicates of 8,192 shots.",
+      "inference": "Estimate replicate mean, variance and best-sample quality under one isolated mitigation mode.",
+      "outcome": "Mean ratio 0.6733; best sample 0.8474; usage 24s.",
+      "learning": "DD-only slightly helped; twirling slightly hurt; the combined stack strongly hurt, so mitigation interactions cannot be assumed additive.",
+      "job_id": "d9gke4bsbqfc73eoovq0",
+      "quantum_seconds": 24,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_depth_and_mitigation.json"
+    },
+    {
+      "id": "E15",
+      "execution": "IBM hardware",
+      "input": "Fixed best p=1 parameters; XpXm dynamical decoupling only; 10 replicates of 8,192 shots.",
+      "inference": "Estimate replicate mean, variance and best-sample quality under one isolated mitigation mode.",
+      "outcome": "Mean ratio 0.6792; best sample 0.8520; usage 24s.",
+      "learning": "DD-only slightly helped; twirling slightly hurt; the combined stack strongly hurt, so mitigation interactions cannot be assumed additive.",
+      "job_id": "d9gkekggk0ls73f1mlsg",
+      "quantum_seconds": 24,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_depth_and_mitigation.json"
+    },
+    {
+      "id": "E16",
+      "execution": "IBM hardware",
+      "input": "Fixed best p=1 parameters; Gate and measurement twirling only; 10 replicates of 8,192 shots.",
+      "inference": "Estimate replicate mean, variance and best-sample quality under one isolated mitigation mode.",
+      "outcome": "Mean ratio 0.6653; best sample 0.8231; usage 25s.",
+      "learning": "DD-only slightly helped; twirling slightly hurt; the combined stack strongly hurt, so mitigation interactions cannot be assumed additive.",
+      "job_id": "d9gkf4t0k0jc738gt5hg",
+      "quantum_seconds": 25,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_depth_and_mitigation.json"
+    },
+    {
+      "id": "E17",
+      "execution": "IBM hardware",
+      "input": "Fixed best p=1 parameters; XpXm DD plus gate/measurement twirling; 10 replicates of 8,192 shots.",
+      "inference": "Estimate replicate mean, variance and best-sample quality under one isolated mitigation mode.",
+      "outcome": "Mean ratio 0.5557; best sample 0.7549; usage 25s.",
+      "learning": "DD-only slightly helped; twirling slightly hurt; the combined stack strongly hurt, so mitigation interactions cannot be assumed additive.",
+      "job_id": "d9gkg04honhs73aboj8g",
+      "quantum_seconds": 25,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_depth_and_mitigation.json"
+    },
+    {
+      "id": "E18",
+      "execution": "IBM hardware",
+      "input": "Best p=5 and p=6 schedules; 22 independent PUBs each; 4,096 shots per PUB.",
+      "inference": "Estimate high-stat replicate means, standard errors and extreme sample quality head-to-head.",
+      "outcome": "p5 mean 0.73663; p6 mean 0.74227; strongest p6 sample 0.91561.",
+      "learning": "The replicated result confirms p=6 as the mean-quality winner and raises the strongest observed sample to 91.56% of exact.",
+      "job_id": "d9gl69t0k0jc738gu9mg",
+      "quantum_seconds": 51,
+      "status": "COMPLETED",
+      "evidence": "outputs/ising150_quota_finale.json"
+    }
+  ],
+  "totals": {
+    "documented_experiments": 19,
+    "successful_hardware_jobs": 15,
+    "quantum_seconds": 599,
+    "allowance_seconds": 600,
+    "remaining_seconds": 1
+  },
+  "failed_workloads": [
+    {
+      "job_id": "d9glaishonhs73abprjg",
+      "status": "ERROR",
+      "usage": 0,
+      "reason": "IBM error 1305: one-second max_execution_time was too small; no result and no billed usage."
+    },
+    {
+      "job_id": "d9glbkshonhs73abpt5g",
+      "status": "CANCELLED",
+      "usage": 0,
+      "reason": "150-qubit 128-shot control remained account-queued with one second left; cancelled unbilled."
+    },
+    {
+      "job_id": "d9gli80gk0ls73f1oae0",
+      "status": "CANCELLED",
+      "usage": 0,
+      "reason": "One-qubit 128-shot control on ibm_fez remained account-queued; cancelled unbilled."
+    },
+    {
+      "job_id": "d9glklggk0ls73f1odu0",
+      "status": "CANCELLED",
+      "usage": 0,
+      "reason": "One-qubit 128-shot control on ibm_kingston also remained account-queued; cancelled unbilled."
+    }
+  ],
+  "claim_boundary": {
+    "hardware_execution": "verified",
+    "exact_classical_baseline": "verified",
+    "quantum_acceleration": "not measured",
+    "quantum_advantage": "not claimed"
+  }
+}

evidence/release_summary.json

@@ -0,0 +1,338 @@
+{
+  "title": "150-Qubit Hardware-Native Ising Optimization Study",
+  "date": "2026-07-23",
+  "backend": "ibm_fez",
+  "processor_qubits": 156,
+  "active_physical_qubits": 150,
+  "problem": {
+    "variables": 150,
+    "search_space": "2^150",
+    "search_space_approx": "1.427e45",
+    "weighted_xor_constraints": 167,
+    "total_weight": 873
+  },
+  "classical_exact": {
+    "solver": "scipy.optimize.milp / HiGHS",
+    "optimal_weight": 865,
+    "total_weight": 873,
+    "fraction": 0.9908361970217641,
+    "mip_gap": 0.0,
+    "wall_seconds": 3.2867870409972966
+  },
+  "hardware_depth_curve": [
+    {
+      "p": 1,
+      "job_id": "d9gke4bsbqfc73eoovq0",
+      "quantum_seconds": 24,
+      "pubs": 10,
+      "shots_per_pub": 8192,
+      "shots": 81920,
+      "two_qubit_gates": 334,
+      "compiled_depth": 381,
+      "mean_approximation_ratio": 0.6733289209266619,
+      "best_sample_approximation_ratio": 0.8473988439306358
+    },
+    {
+      "p": 2,
+      "job_id": "d9gk9kjsbqfc73eoops0",
+      "quantum_seconds": 57,
+      "pubs": 50,
+      "shots_per_pub": 4096,
+      "shots": 204800,
+      "two_qubit_gates": 668,
+      "compiled_depth": 406,
+      "mean_approximation_ratio": 0.6903317490065028,
+      "best_sample_approximation_ratio": 0.8208092485549133
+    },
+    {
+      "p": 3,
+      "job_id": "d9gkajl0k0jc738gsv80",
+      "quantum_seconds": 57,
+      "pubs": 50,
+      "shots_per_pub": 4096,
+      "shots": 204800,
+      "two_qubit_gates": 1002,
+      "compiled_depth": 431,
+      "mean_approximation_ratio": 0.7048139450867053,
+      "best_sample_approximation_ratio": 0.8439306358381503
+    },
+    {
+      "p": 4,
+      "job_id": "d9gkj98gk0ls73f1mspg",
+      "quantum_seconds": 57,
+      "pubs": 50,
+      "shots_per_pub": 4096,
+      "shots": 204800,
+      "two_qubit_gates": 1336,
+      "compiled_depth": 456,
+      "mean_approximation_ratio": 0.7262520321531792,
+      "best_sample_approximation_ratio": 0.8578034682080925
+    },
+    {
+      "p": 5,
+      "job_id": "d9gkle3sbqfc73eop9tg",
+      "quantum_seconds": 58,
+      "pubs": 50,
+      "shots_per_pub": 4096,
+      "shots": 204800,
+      "two_qubit_gates": 1670,
+      "compiled_depth": 481,
+      "mean_approximation_ratio": 0.7356699331647398,
+      "best_sample_approximation_ratio": 0.892485549132948
+    },
+    {
+      "p": 6,
+      "job_id": "d9gknv4honhs73abouig",
+      "quantum_seconds": 58,
+      "pubs": 50,
+      "shots_per_pub": 4096,
+      "shots": 204800,
+      "two_qubit_gates": 2004,
+      "compiled_depth": 506,
+      "mean_approximation_ratio": 0.739049516799133,
+      "best_sample_approximation_ratio": 0.8774566473988439
+    },
+    {
+      "p": 7,
+      "job_id": "d9gkq3ogk0ls73f1n5tg",
+      "quantum_seconds": 58,
+      "pubs": 50,
+      "shots_per_pub": 4096,
+      "shots": 204800,
+      "two_qubit_gates": 2338,
+      "compiled_depth": 531,
+      "mean_approximation_ratio": 0.736153416275289,
+      "best_sample_approximation_ratio": 0.8647398843930636
+    },
+    {
+      "p": 8,
+      "job_id": "d9gkrnjsbqfc73eopi3g",
+      "quantum_seconds": 58,
+      "pubs": 50,
+      "shots_per_pub": 4096,
+      "shots": 204800,
+      "two_qubit_gates": 2672,
+      "compiled_depth": 556,
+      "mean_approximation_ratio": 0.7354015195989885,
+      "best_sample_approximation_ratio": 0.8797687861271676
+    }
+  ],
+  "mitigation_ablation": [
+    {
+      "mode": "raw",
+      "job_id": "d9gke4bsbqfc73eoovq0",
+      "quantum_seconds": 24,
+      "pubs": 10,
+      "shots": 81920,
+      "mean_approximation_ratio": 0.6733289209266619,
+      "best_sample_approximation_ratio": 0.8473988439306358
+    },
+    {
+      "mode": "dd_only",
+      "job_id": "d9gkekggk0ls73f1mlsg",
+      "quantum_seconds": 24,
+      "pubs": 10,
+      "shots": 81920,
+      "mean_approximation_ratio": 0.6792395654579119,
+      "best_sample_approximation_ratio": 0.8520231213872832
+    },
+    {
+      "mode": "twirling_only",
+      "job_id": "d9gkf4t0k0jc738gt5hg",
+      "quantum_seconds": 25,
+      "pubs": 10,
+      "shots": 81920,
+      "mean_approximation_ratio": 0.6653075042901012,
+      "best_sample_approximation_ratio": 0.823121387283237
+    },
+    {
+      "mode": "dd_plus_twirling",
+      "job_id": "d9gkg04honhs73aboj8g",
+      "quantum_seconds": 25,
+      "pubs": 10,
+      "shots": 81920,
+      "mean_approximation_ratio": 0.5556691287707731,
+      "best_sample_approximation_ratio": 0.7549132947976879
+    }
+  ],
+  "jobs": [
+    {
+      "role": "original_smoke",
+      "job_id": "d9gjobkhonhs73abnl3g",
+      "quantum_seconds": 3,
+      "pubs": 3,
+      "shots": 3072
+    },
+    {
+      "role": "p1_parameter_grid",
+      "job_id": "d9gk5gchonhs73abo5ag",
+      "quantum_seconds": 30,
+      "pubs": 100,
+      "shots": 102400
+    },
+    {
+      "role": "initial_mitigated_confirmation",
+      "job_id": "d9gk64ogk0ls73f1mar0",
+      "quantum_seconds": 14,
+      "pubs": 10,
+      "shots": 40960
+    },
+    {
+      "role": "p2_depth",
+      "job_id": "d9gk9kjsbqfc73eoops0",
+      "quantum_seconds": 57,
+      "pubs": 50,
+      "shots": 204800
+    },
+    {
+      "role": "p3_depth",
+      "job_id": "d9gkajl0k0jc738gsv80",
+      "quantum_seconds": 57,
+      "pubs": 50,
+      "shots": 204800
+    },
+    {
+      "role": "p4_depth",
+      "job_id": "d9gkj98gk0ls73f1mspg",
+      "quantum_seconds": 57,
+      "pubs": 50,
+      "shots": 204800
+    },
+    {
+      "role": "p5_depth",
+      "job_id": "d9gkle3sbqfc73eop9tg",
+      "quantum_seconds": 58,
+      "pubs": 50,
+      "shots": 204800
+    },
+    {
+      "role": "p6_depth",
+      "job_id": "d9gknv4honhs73abouig",
+      "quantum_seconds": 58,
+      "pubs": 50,
+      "shots": 204800
+    },
+    {
+      "role": "p7_depth",
+      "job_id": "d9gkq3ogk0ls73f1n5tg",
+      "quantum_seconds": 58,
+      "pubs": 50,
+      "shots": 204800
+    },
+    {
+      "role": "p8_depth",
+      "job_id": "d9gkrnjsbqfc73eopi3g",
+      "quantum_seconds": 58,
+      "pubs": 50,
+      "shots": 204800
+    },
+    {
+      "role": "ablation_raw",
+      "job_id": "d9gke4bsbqfc73eoovq0",
+      "quantum_seconds": 24,
+      "pubs": 10,
+      "shots": 81920
+    },
+    {
+      "role": "ablation_dd_only",
+      "job_id": "d9gkekggk0ls73f1mlsg",
+      "quantum_seconds": 24,
+      "pubs": 10,
+      "shots": 81920
+    },
+    {
+      "role": "ablation_twirling_only",
+      "job_id": "d9gkf4t0k0jc738gt5hg",
+      "quantum_seconds": 25,
+      "pubs": 10,
+      "shots": 81920
+    },
+    {
+      "role": "ablation_dd_plus_twirling",
+      "job_id": "d9gkg04honhs73aboj8g",
+      "quantum_seconds": 25,
+      "pubs": 10,
+      "shots": 81920
+    }
+  ],
+  "totals": {
+    "jobs": 15,
+    "qaoa_pubs": 544,
+    "all_pubs": 547,
+    "qaoa_shots": 2084864,
+    "all_shots": 2087936,
+    "quantum_seconds": 599,
+    "allowance_seconds": 600,
+    "remaining_seconds": 1
+  },
+  "findings": {
+    "best_mean_depth": 6,
+    "best_mean_approximation_ratio": 0.739049516799133,
+    "best_sample_depth": 5,
+    "best_sample_approximation_ratio": 0.892485549132948,
+    "dd_only_delta_vs_raw": 0.005910644531250009,
+    "twirling_only_delta_vs_raw": -0.008021416636560641,
+    "combined_delta_vs_raw": -0.11765979215588873,
+    "replicated_best_mean_depth": 6,
+    "replicated_best_mean_approximation_ratio": 0.742274123095113,
+    "replicated_best_sample_depth": 6,
+    "replicated_best_sample_approximation_ratio": 0.915606936416185
+  },
+  "truth_boundary": {
+    "hardware_execution_verified": true,
+    "classical_exact_solution_verified": true,
+    "quantum_acceleration_measured": false,
+    "quantum_advantage_claimed": false,
+    "note": "Classical HiGHS solved this instance exactly and remains the production winner."
+  },
+  "final_p5_p6_confirmation": {
+    "job_id": "d9gl69t0k0jc738gu9mg",
+    "usage": 51,
+    "pubs": 44,
+    "shots": 180224,
+    "groups": {
+      "p5": {
+        "replicates": 22,
+        "total_shots": 90112,
+        "mean_approximation_ratio": 0.7366256353044535,
+        "standard_deviation_across_replicates": 0.0007774597290066907,
+        "standard_error": 0.00016575497116186523,
+        "best_sample_approximation_ratio": 0.9109826589595376
+      },
+      "p6": {
+        "replicates": 22,
+        "total_shots": 90112,
+        "mean_approximation_ratio": 0.742274123095113,
+        "standard_deviation_across_replicates": 0.00071458200975097,
+        "standard_error": 0.0001523493963737386,
+        "best_sample_approximation_ratio": 0.915606936416185
+      }
+    }
+  },
+  "failed_workloads": [
+    {
+      "job_id": "d9glaishonhs73abprjg",
+      "status": "ERROR",
+      "usage": 0,
+      "reason": "IBM error 1305: one-second max_execution_time was too small; no result and no billed usage."
+    },
+    {
+      "job_id": "d9glbkshonhs73abpt5g",
+      "status": "CANCELLED",
+      "usage": 0,
+      "reason": "150-qubit 128-shot control remained account-queued with one second left; cancelled unbilled."
+    },
+    {
+      "job_id": "d9gli80gk0ls73f1oae0",
+      "status": "CANCELLED",
+      "usage": 0,
+      "reason": "One-qubit 128-shot control on ibm_fez remained account-queued; cancelled unbilled."
+    },
+    {
+      "job_id": "d9glklggk0ls73f1odu0",
+      "status": "CANCELLED",
+      "usage": 0,
+      "reason": "One-qubit 128-shot control on ibm_kingston also remained account-queued; cancelled unbilled."
+    }
+  ]
+}

evidence/source_manifest.json

@@ -0,0 +1,93 @@
+{
+  "files": {
+    "README.md": {
+      "sha256": "7dfea5ad425b63af022b9b1747c8d4e7ebd7e182a93761e6d83a04f7f11175b4",
+      "bytes": 4351
+    },
+    "EXPERIMENTS.md": {
+      "sha256": "cc0f6485fcb9930e5145527081ded6142f0a85c527b7cd0f3554da814c2628c6",
+      "bytes": 13325
+    },
+    "LEARNINGS_AND_NEXT_EXPERIMENTS.md": {
+      "sha256": "6917c412792087f9ca0e2f753fecf659b51f2a3ce9c5b7601387fc473de9a63a",
+      "bytes": 8593
+    },
+    "references/bouland-papers.bib": {
+      "sha256": "f3d8cd2fae07df2b6dab0df65c5bf70396d744017126b853f379d61cede1f4ad",
+      "bytes": 1998
+    },
+    "next_experiments/__init__.py": {
+      "sha256": "98f832a59acc5f9402daf0e7ac87668811dd8eb1470e26ca3535e8309d148350",
+      "bytes": 412
+    },
+    "next_experiments/rcs_metrics.py": {
+      "sha256": "539f600eb185e00d72d10148a5351d94349716649415c0508fa4b119809ff5f2",
+      "bytes": 4192
+    },
+    "next_experiments/run_local_rcs.py": {
+      "sha256": "380a57d7d80f3f5ec45e09311edd48953ea60ae9807ed32723d67be2ad9991e5",
+      "bytes": 3938
+    },
+    "next_experiments/validate_series.py": {
+      "sha256": "2ad5807698fb2204c398a73d778f27ca340c8c64fe65c8d18ac6070c66fb15e8",
+      "bytes": 3475
+    },
+    "next_experiments/manifests/bouland-series.json": {
+      "sha256": "95b82d0a3b54f50c0be71a7c8ffcc50bdbdedccc81013823a5b1bdb88a2063ee",
+      "bytes": 7096
+    },
+    "outputs/experiment_ledger.json": {
+      "sha256": "7ca784b9848929e0b36ede0337059495b20c9ed06ba38920fed034622c4ba49c",
+      "bytes": 13873
+    },
+    "outputs/ising150_release_summary.json": {
+      "sha256": "4c4de5f47b5da6cd72e34b19315f7262ac3b6f1f6b12fc7908e1fa3f5a8b9cc3",
+      "bytes": 9263
+    },
+    "outputs/ising150_depth_curve.csv": {
+      "sha256": "0926203aae5d00441ef631aeab5f471d546a131ada3b39323c47ccbda2286d09",
+      "bytes": 845
+    },
+    "outputs/ising150_quota_finale.json": {
+      "sha256": "5eb151a58b000c3b07e602711875bfd9c385f5dbcc9acbb2c2715bce904c26e3",
+      "bytes": 125688
+    },
+    "outputs/kaggle_ibm_quantum_spectacle.ipynb": {
+      "sha256": "267c03f82aa41fd5005e49b0d858163a8fc3c5ee79e6b207980d6c2b9260d6eb",
+      "bytes": 11650
+    },
+    "outputs/X_POST_DRAFT.md": {
+      "sha256": "81b8548eccf8c602818c06fd623a1d719cb8567c543152ecd9341ca7a013bc05",
+      "bytes": 964
+    },
+    "outputs/next_experiments/local_rcs_smoke.json": {
+      "sha256": "fad4bfdd42c5e1c5b8d920cf81c19a02bdbb1cba894bb44da9371b9c2d90881b",
+      "bytes": 1056
+    },
+    "outputs/ising150_report.html": {
+      "sha256": "6e2c77816a0c34a59233d109b5ae90a9ca9b9ed6e4094e8fedcb6e9f42a955e9",
+      "bytes": 7278
+    },
+    "outputs/ising150_recording.mp4": {
+      "sha256": "1dd46b4f6abdc47afb38205af05e080bb2720fe82805ca1e8c2d1cd512c60fa3",
+      "bytes": 904048
+    },
+    "outputs/ising150_recording.gif": {
+      "sha256": "6bb447b3cd1963ea72c6561e5c1cc29869af2e0b106083a7b97a89fae5a64083",
+      "bytes": 1574379
+    }
+  },
+  "totals": {
+    "documented_experiments": 19,
+    "successful_hardware_jobs": 15,
+    "quantum_seconds": 599,
+    "allowance_seconds": 600,
+    "remaining_seconds": 1
+  },
+  "claim_boundary": {
+    "hardware_execution": "verified",
+    "exact_classical_baseline": "verified",
+    "quantum_acceleration": "not measured",
+    "quantum_advantage": "not claimed"
+  }
+}