Dataset Viewer
Auto-converted to Parquet Duplicate
text
stringlengths
0
262
============================= test session starts ==============================
platform linux -- Python 3.12.13, pytest-8.4.1, pluggy-1.6.0 -- /usr/local/bin/python3
rootdir: /app
configfile: ../dev/null
plugins: json-ctrf-0.3.5
collecting ... collected 14 items
::test_frozen_instance_untampered PASSED [ 7%]
::test_results_present_and_complete PASSED [ 14%]
::test_solver_is_a_real_computation PASSED [ 21%]
::test_each_config_matches_independent_recompute FAILED [ 28%]
::test_reciprocity_and_energy_invariants PASSED [ 35%]
::test_hybridisation_widens_the_gap PASSED [ 42%]
::test_ratio_ordering_across_configs PASSED [ 50%]
::test_effective_index_follows_compliance_law PASSED [ 57%]
::test_held_out_solver_generalises FAILED [ 64%]
::test_diatomic_solver_is_a_real_computation PASSED [ 71%]
::test_diatomic_configs_match_independent_recompute FAILED [ 78%]
::test_diatomic_reciprocity_and_energy PASSED [ 85%]
::test_diatomic_held_out_solver_generalises FAILED [ 92%]
::test_no_leaked_reference_artifacts PASSED [100%]
=================================== FAILURES ===================================
________________ test_each_config_matches_independent_recompute ________________
def test_each_config_matches_independent_recompute():
LOGS.mkdir(parents=True, exist_ok=True)
summary = []
for entry, rec, ref in _iter_configs():
where = f"config {entry['id']}"
assert int(rec["N"]) == int(entry["N"]) if "N" in rec else True
> grade_config(rec, ref, where)
/verifier/test_outputs.py:853:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
/verifier/test_outputs.py:664: in grade_config
grade_scalar(rec, ref, "gap_decay_rate", 0.08, 0.02, where)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
rec = {'N': 12, 'TL_peak_bragg': [18.91333813716943, 248.94174546730915], 'alpha_max': 0.0, 'attn_bragg_max': 0.23778965375369304, ...}
ref = {'TL_peak_bragg': [18.913343539344712, 248.96229019522488], '_N': 12, '_fB': 265.06094856272057, '_fH': 441.76850207839675, ...}
key = 'gap_decay_rate', rel = 0.08, ab = 0.02, where = 'config detuned_below'
def grade_scalar(rec, ref, key, rel, ab, where):
assert key in rec and rec[key] is not None, f"{where}: missing '{key}'"
> assert approx(rec[key], ref[key], rel, ab), (
f"{where}: {key}={rec[key]} vs reference {ref[key]} (rel={rel}, abs={ab})")
E AssertionError: config detuned_below: gap_decay_rate=-0.22222314644990812 vs reference 0.22230924873449445 (rel=0.08, abs=0.02)
E assert False
E + where False = approx(-0.22222314644990812, 0.22230924873449445, 0.08, 0.02)
/verifier/test_outputs.py:596: AssertionError
_______________________ test_held_out_solver_generalises _______________________
def test_held_out_solver_generalises():
"""Run the agent's own solver on hidden geometries and grade against the
independent recompute. Hardcoded / instance-specific solvers fail here."""
assert SOLVER.is_file(), "Missing solver"
LOGS.mkdir(parents=True, exist_ok=True)
inst = load_instance()
const = inst["constants"]
records = []
for label, spec in HIDDEN_SPECS:
p = _params_from_ratio(spec, const["c"], const["rho"], const["r_d"])
cmd = [sys.executable, str(SOLVER),
"--r_n", repr(p["r_n"]), "--l_n", repr(p["l_n"]), "--V", repr(p["V"]),
"--d", repr(p["d"]), "--N", str(p["N"]),
"--Rv_coeff", repr(p["Rv_coeff"]),
"--r_d", repr(p["r_d"]), "--c", repr(p["c"]), "--rho", repr(p["rho"])]
if not p["has_cavity"]:
cmd.append("--no-cavity")
proc = subprocess.run(cmd, capture_output=True, text=True, timeout=600)
assert proc.returncode == 0, f"held-out {label}: solver failed: {proc.stderr[-800:]}"
payload = load_json_from_stdout(proc.stdout)
ref = ref_solve(tuple(sorted(p.items())))
> grade_config(payload, ref, f"HELD-OUT {label}")
/verifier/test_outputs.py:952:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
/verifier/test_outputs.py:664: in grade_config
grade_scalar(rec, ref, "gap_decay_rate", 0.08, 0.02, where)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
rec = {'N': 12, 'TL_peak_bragg': [42.21196996534466, 304.2948961613768], 'alpha_max': 0.0, 'attn_bragg_max': 0.45618925980248065, ...}
ref = {'TL_peak_bragg': [42.211968202109446, 304.2806552316506], '_N': 12, '_fB': 334.32789832481154, '_fH': 417.9098729060144, ...}
key = 'gap_decay_rate', rel = 0.08, ab = 0.02
where = 'HELD-OUT holdout_partial'
def grade_scalar(rec, ref, key, rel, ab, where):
assert key in rec and rec[key] is not None, f"{where}: missing '{key}'"
> assert approx(rec[key], ref[key], rel, ab), (
f"{where}: {key}={rec[key]} vs reference {ref[key]} (rel={rel}, abs={ab})")
E AssertionError: HELD-OUT holdout_partial: gap_decay_rate=-0.4526354491298243 vs reference 0.4540500924209423 (rel=0.08, abs=0.02)
E assert False
E + where False = approx(-0.4526354491298243, 0.4540500924209423, 0.08, 0.02)
/verifier/test_outputs.py:596: AssertionError
______________ test_diatomic_configs_match_independent_recompute _______________
def test_diatomic_configs_match_independent_recompute():
End of preview. Expand in Data Studio

No dataset card yet

Downloads last month
52