The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/harbor/harbor.py", line 171, in _split_generators
raise DataFilesNotFoundError("No task.toml or instruction.md files found")
datasets.exceptions.DataFilesNotFoundError: No task.toml or instruction.md files found
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
OSC exact two carrier reference
Download the complete offline reproducibility package and archive SHA256. Extract the ZIP and run the commands below inside its osc-method-submission/ directory. The archive includes all code, source-data snapshots, proof, results, controls, and the internal file manifest.
Archive SHA256: d888f4c1d34982f6bb6678ad855f1960923224baa7233fd7a33983981ca8e9b6.
Human entrant and contact: Riley Sturgill (rsturg). AI assistance: OpenAI Codex assisted with research, mathematical derivation, implementation and local verification.
This offline reproducibility package proves and calculates a zero d-wave ground-state reference for a specific dilute sector of the periodic square Hubbard model. The primary case is 6×6, one up and one down electron, U/t = 8. Its exact maximum-distance d-wave correlation is zero. Full-sector and relative-coordinate diagonalization independently give E₀/t = −7.906146146983354.
It is prepared as a candidate for the Open Superconductor Challenge's $1,000 Best Method prize, subject to organizer reproduction and comparative review. The package is frozen for public Method review; organizer acceptance and any award remain pending. A positive correlation in the organizer's biased variational state is not contradicted by a zero true-ground-state reference. (Challenge)
Reproduce the calculation
Tested environment: Python 3.11.1 and NumPy 2.4.6; requirements.txt pins NumPy. Only NumPy and Python standard-library modules are used. Once NumPy is installed, no GPU, model checkpoint, network connection or Monte Carlo training is required. The supplied runners verify the frozen-file SHA256 manifest before computing:
sh run_submission.sh
On Windows with the Python launcher:
.\run_submission.ps1
The equivalent direct calculation from the extracted package directory is:
python osc_exact_two.py --out results_recomputed.json --active-csv sources/active_challenge.csv --affected-out affected_materials_recomputed.json
If the frozen active CSV is elsewhere, replace that path. To run the physics checks without the table audit:
python osc_exact_two.py --out results_recomputed.json
The script freshly constructs and diagonalizes all Hamiltonians; it does not load cached ground states or fit published amplitudes. It exits with an assertion failure if a check fails. The console prints the 6×6 result, and the JSON contains all cases, controls and the exact-rational energy bracket. Runtime and last floating-point digits vary between numerical libraries. Recomputed files need not be byte-identical to the recorded JSON because runtime measurements are included.
What is proved and checked
METHOD.md contains the mathematical proof. Uniform t > 0, finite U, a flux-free periodic square lattice and the zero-temperature one-up/one-down sector are essential assumptions. Perron–Frobenius and quarter-turn symmetry force each anchored d-wave annihilation amplitude to cancel. Every pair correlation then vanishes exactly.
On the even 6×6 lattice, the same maximum-distance zero follows for 70 electrons by particle-hole symmetry only where the two pair-operator supports are disjoint. The onsite two-hole d-wave correlation is nonzero. The implementation checks this extension with actual electron creation/annihilation signs rather than reusing the two-electron formula.
The numerical checks include full 1,296-state versus separate 36-state diagonalization; L = 3, 4 and 6; U/t = 0 and 8; t = 0.37 scale invariance; explicit fermion signs on a random complex state; a nonzero s-wave control; a nonzero anisotropic d-wave control; and exact Fraction sign checks bounding E₀/t between −7.906146146984 and −7.906146146982. The only random seed, 20261025, belongs to the operator probe. Ground-state calculations are deterministic.
Frozen material example and data audit
The primary example is 1FeH2O2-1 / OSC-00931: t = 0.18777 eV, U = 1.50216 eV, N(E_F) = 0.60844, and published effective solver_delta = 0.9444. Its published amplitude 0.00283 gives proxy index 0.51656556 at scale 300. The proposed exact-ground-state reference gives amplitude and reference index zero; it does not change the canonical leaderboard.
The frozen table contains 1,034 active rows. The audit selects 86 rows sharing that effective dilute-sector amplitude: 19 with n_band < 1 and 67 above 1, interpreted through the public particle-hole pipeline. The latter interpretation is conditional on that convention; solver_Ne is absent from the CSV header and ph_mapped is blank for these rows. The package leaves all DFT/downfolding inputs unchanged. Data revision: 4d46054a0a7c65344b5b3af9ed2e365829f184b3. The CSV hash is recorded in affected_materials.json.
Files and provenance
osc_exact_two.py is the original calculation and validation code. METHOD.md supplies the proof, numerical interpretation and source links. results.json and affected_materials.json record the local run. sources/ preserves selected official inputs and attributed references; sources/reference_metadata.json identifies other primary sources by URL and hash. submission_post.md contains the prepared Method-review request; submission.json is the instrument-schema ground-state reference for OSC-00931. verify_manifest.py checks the final SHA256 manifest. ATTRIBUTION.md identifies third-party terms and prior work.
Original method code is MIT licensed as stated in its source. Third-party material data and reference documents retain CC-BY-NC 4.0, C2DB/FINAL-Bench attribution and their notices. Ku-ku's earlier validation and Tuli99's separate Method entry are acknowledged; no global novelty or first-ever ED claim is made. Read the preserved provenance before redistributing source data.
Submission status and remaining review
The latest organizer clarification makes the Hugging Face dataset discussion board the official Method handoff channel; no separate email is required. An older README email instruction is superseded by that explicit clarification. A new Method thread should supply the reviewed package and real entrant identity; sign-in alone is not organizer verification. (Canonical organizer rules, explicit delivery clarification)
The organizer must confirm the private 6×6 boundary and observable conventions and decide whether this ground-state calibration is an admissible Method contribution. The human entrant is Riley Sturgill (rsturg), who has authorized publication with explicit OpenAI Codex disclosure. After an author is supplied, the code can produce a local instrument-schema JSON with --author, --active-csv and --submission-out; this never transmits data. Account setup and authorized publication are separate from the offline computation; organizer acceptance and any award remain pending.
- Downloads last month
- 47