Neural Solver Synthesis
Collection
Official artifacts for Neural Solver Synthesis: models, datasets, certified evaluations, cross-domain results, and final evidence. • 55 items • Updated
Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code: DatasetGenerationCastError
Exception: DatasetGenerationCastError
Message: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 16 new columns ({'CertifiedOptimalGapLowerMean', 'CertifiedOptimalReferenceRelativeIntervalMax', 'CertifiedOptimumAttainmentMean', 'CertifiedOptimalGapUpperMean', 'PassRateFeasibleBenchmarkMean', 'PassRateFeasibleBenchmarkStd', 'AllFeasibleGapUpperMean', 'FalseFeasibleOnProvedInfeasible', 'CertifiedOptimalGapMean', 'AllFeasibleGapLowerMean', 'CertifiedOptimalGapStd', 'Method', 'ProvedInfeasibleRows', 'CertifiedOptimalReferenceIntervalMax', 'CertifiedOptimalReferenceIntervalMean', 'Seeds'}) and 24 missing columns ({'FeasibleBenchmarkRows', 'PairedCertifiedGapDifferenceUpperReference', 'PairedPassDifference', 'PairedCertifiedGapDifference', 'ComparatorPassRate', 'BootstrapSamples', 'ComparatorCertifiedGap', 'BootstrapSeed', 'PairedCertifiedGapDifferenceCI025', 'PairedPassDifferenceCI975', 'ReferenceCertifiedGap', 'ComparatorCertifiedGapUpper', 'ComparatorCertifiedGapLower', 'ReferenceGapWins', 'ReferenceMethod', 'ReferencePassRate', 'CertifiedGapTies', 'ReferenceCertifiedGapUpper', 'ComparatorGapWins', 'ComparatorMethod', 'ReferenceCertifiedGapLower', 'PairedPassDifferenceCI025', 'CertifiedOptimalRows', 'PairedCertifiedGapDifferenceCI975'}).
This happened while the json dataset builder was generating data using
hf://datasets/IDEALLab/Neural-Solver-Synthesis-Final-Evidence-v1/artifacts/neurips2026/hypothesize/training_step90_summary.json (at revision 9689ee9ce844f904a54b6736123662499ba26681), ['hf://datasets/IDEALLab/Neural-Solver-Synthesis-Final-Evidence-v1@9689ee9ce844f904a54b6736123662499ba26681/artifacts/neurips2026/hypothesize/training_step90_paired.json', 'hf://datasets/IDEALLab/Neural-Solver-Synthesis-Final-Evidence-v1@9689ee9ce844f904a54b6736123662499ba26681/artifacts/neurips2026/hypothesize/training_step90_summary.json'], ['hf://datasets/IDEALLab/Neural-Solver-Synthesis-Final-Evidence-v1@9689ee9ce844f904a54b6736123662499ba26681/artifacts/neurips2026/hypothesize/training_step90_paired.json', 'hf://datasets/IDEALLab/Neural-Solver-Synthesis-Final-Evidence-v1@9689ee9ce844f904a54b6736123662499ba26681/artifacts/neurips2026/hypothesize/training_step90_summary.json']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1848, in _prepare_split_single
writer.write_table(table)
~~~~~~~~~~~~~~~~~~^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 765, in write_table
self._write_table(pa_table, writer_batch_size=writer_batch_size)
~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
pa_table = table_cast(pa_table, self._schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
AllFeasibleGapLowerMean: double
AllFeasibleGapUpperMean: double
CertifiedOptimalGapLowerMean: double
CertifiedOptimalGapMean: double
CertifiedOptimalGapStd: double
CertifiedOptimalGapUpperMean: double
CertifiedOptimalReferenceIntervalMax: double
CertifiedOptimalReferenceIntervalMean: double
CertifiedOptimalReferenceRelativeIntervalMax: double
CertifiedOptimumAttainmentMean: double
FalseFeasibleOnProvedInfeasible: int64
Method: string
PassRateFeasibleBenchmarkMean: double
PassRateFeasibleBenchmarkStd: double
ProvedInfeasibleRows: int64
Seeds: int64
to
{'BootstrapSamples': Value('int64'), 'BootstrapSeed': Value('int64'), 'CertifiedGapTies': Value('int64'), 'CertifiedOptimalRows': Value('int64'), 'ComparatorCertifiedGap': Value('float64'), 'ComparatorCertifiedGapLower': Value('float64'), 'ComparatorCertifiedGapUpper': Value('float64'), 'ComparatorGapWins': Value('int64'), 'ComparatorMethod': Value('string'), 'ComparatorPassRate': Value('float64'), 'FeasibleBenchmarkRows': Value('int64'), 'PairedCertifiedGapDifference': Value('float64'), 'PairedCertifiedGapDifferenceCI025': Value('float64'), 'PairedCertifiedGapDifferenceCI975': Value('float64'), 'PairedCertifiedGapDifferenceUpperReference': Value('float64'), 'PairedPassDifference': Value('float64'), 'PairedPassDifferenceCI025': Value('float64'), 'PairedPassDifferenceCI975': Value('float64'), 'ReferenceCertifiedGap': Value('float64'), 'ReferenceCertifiedGapLower': Value('float64'), 'ReferenceCertifiedGapUpper': Value('float64'), 'ReferenceGapWins': Value('int64'), 'ReferenceMethod': Value('string'), 'ReferencePassRate': Value('float64')}
because column names don't match
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1850, in _prepare_split_single
raise DatasetGenerationCastError.from_cast_error(
...<4 lines>...
)
datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 16 new columns ({'CertifiedOptimalGapLowerMean', 'CertifiedOptimalReferenceRelativeIntervalMax', 'CertifiedOptimumAttainmentMean', 'CertifiedOptimalGapUpperMean', 'PassRateFeasibleBenchmarkMean', 'PassRateFeasibleBenchmarkStd', 'AllFeasibleGapUpperMean', 'FalseFeasibleOnProvedInfeasible', 'CertifiedOptimalGapMean', 'AllFeasibleGapLowerMean', 'CertifiedOptimalGapStd', 'Method', 'ProvedInfeasibleRows', 'CertifiedOptimalReferenceIntervalMax', 'CertifiedOptimalReferenceIntervalMean', 'Seeds'}) and 24 missing columns ({'FeasibleBenchmarkRows', 'PairedCertifiedGapDifferenceUpperReference', 'PairedPassDifference', 'PairedCertifiedGapDifference', 'ComparatorPassRate', 'BootstrapSamples', 'ComparatorCertifiedGap', 'BootstrapSeed', 'PairedCertifiedGapDifferenceCI025', 'PairedPassDifferenceCI975', 'ReferenceCertifiedGap', 'ComparatorCertifiedGapUpper', 'ComparatorCertifiedGapLower', 'ReferenceGapWins', 'ReferenceMethod', 'ReferencePassRate', 'CertifiedGapTies', 'ReferenceCertifiedGapUpper', 'ComparatorGapWins', 'ComparatorMethod', 'ReferenceCertifiedGapLower', 'PairedPassDifferenceCI025', 'CertifiedOptimalRows', 'PairedCertifiedGapDifferenceCI975'}).
This happened while the json dataset builder was generating data using
hf://datasets/IDEALLab/Neural-Solver-Synthesis-Final-Evidence-v1/artifacts/neurips2026/hypothesize/training_step90_summary.json (at revision 9689ee9ce844f904a54b6736123662499ba26681), ['hf://datasets/IDEALLab/Neural-Solver-Synthesis-Final-Evidence-v1@9689ee9ce844f904a54b6736123662499ba26681/artifacts/neurips2026/hypothesize/training_step90_paired.json', 'hf://datasets/IDEALLab/Neural-Solver-Synthesis-Final-Evidence-v1@9689ee9ce844f904a54b6736123662499ba26681/artifacts/neurips2026/hypothesize/training_step90_summary.json'], ['hf://datasets/IDEALLab/Neural-Solver-Synthesis-Final-Evidence-v1@9689ee9ce844f904a54b6736123662499ba26681/artifacts/neurips2026/hypothesize/training_step90_paired.json', 'hf://datasets/IDEALLab/Neural-Solver-Synthesis-Final-Evidence-v1@9689ee9ce844f904a54b6736123662499ba26681/artifacts/neurips2026/hypothesize/training_step90_summary.json']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)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.
BootstrapSamples int64 | BootstrapSeed int64 | CertifiedGapTies int64 | CertifiedOptimalRows int64 | ComparatorCertifiedGap float64 | ComparatorCertifiedGapLower float64 | ComparatorCertifiedGapUpper float64 | ComparatorGapWins int64 | ComparatorMethod string | ComparatorPassRate float64 | FeasibleBenchmarkRows int64 | PairedCertifiedGapDifference float64 | PairedCertifiedGapDifferenceCI025 float64 | PairedCertifiedGapDifferenceCI975 float64 | PairedCertifiedGapDifferenceUpperReference float64 | PairedPassDifference float64 | PairedPassDifferenceCI025 float64 | PairedPassDifferenceCI975 float64 | ReferenceCertifiedGap float64 | ReferenceCertifiedGapLower float64 | ReferenceCertifiedGapUpper float64 | ReferenceGapWins int64 | ReferenceMethod string | ReferencePassRate float64 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
10,000 | 260,726 | 38 | 2,537 | 0.692754 | 0.692754 | 0.692754 | 124 | Adaptive Repair (64 completions) | 0.59214 | 2,621 | -0.645871 | -0.660019 | -0.633342 | -0.645871 | 0.406715 | 0.387503 | 0.424317 | 0.046883 | 0.046883 | 0.046883 | 2,375 | Historical exact Hero control | 0.998855 |
10,000 | 260,726 | 454 | 2,537 | 0.318068 | 0.318068 | 0.318068 | 496 | Base (Best-of-64) | 0.86074 | 2,621 | -0.271185 | -0.285695 | -0.256596 | -0.271185 | 0.138115 | 0.124957 | 0.151759 | 0.046883 | 0.046883 | 0.046883 | 1,587 | Historical exact Hero control | 0.998855 |
10,000 | 260,726 | 764 | 2,537 | 0.177927 | 0.177927 | 0.177927 | 758 | BnB | 0.988936 | 2,621 | -0.131044 | -0.142172 | -0.119922 | -0.131044 | 0.00992 | 0.006111 | 0.013774 | 0.046883 | 0.046883 | 0.046883 | 1,015 | Historical exact Hero control | 0.998855 |
10,000 | 260,726 | 1,287 | 2,537 | 0 | 0 | 0 | 1,250 | CP-SAT | 1 | 2,621 | 0.046883 | 0.042605 | 0.051123 | 0.046882 | -0.001145 | -0.002647 | 0 | 0.046883 | 0.046883 | 0.046883 | 0 | Historical exact Hero control | 0.998855 |
10,000 | 260,726 | 1,154 | 2,537 | 0.034668 | 0.034668 | 0.034668 | 817 | Frozen Hero | 0.99504 | 2,621 | 0.012215 | 0.00763 | 0.01684 | 0.012215 | 0.003815 | 0.001163 | 0.006941 | 0.046883 | 0.046883 | 0.046883 | 566 | Historical exact Hero control | 0.998855 |
10,000 | 260,726 | 402 | 2,537 | 0.249948 | 0.249948 | 0.249948 | 458 | Greedy | 0.999618 | 2,621 | -0.203065 | -0.214079 | -0.19215 | -0.203065 | -0.000763 | -0.002271 | 0.000759 | 0.046883 | 0.046883 | 0.046883 | 1,677 | Historical exact Hero control | 0.998855 |
10,000 | 260,726 | 1,172 | 2,537 | 0.027674 | 0.027674 | 0.027674 | 1,072 | Hand-written SA | 0.969477 | 2,621 | 0.019208 | 0.014255 | 0.024047 | 0.019208 | 0.029378 | 0.022929 | 0.036219 | 0.046883 | 0.046883 | 0.046883 | 293 | Historical exact Hero control | 0.998855 |
10,000 | 260,726 | 619 | 2,537 | 0.227896 | 0.227896 | 0.227896 | 658 | Historical exact Hero without Hypothesize | 0.819535 | 2,621 | -0.181013 | -0.191158 | -0.166909 | -0.181013 | 0.179321 | 0.165612 | 0.190507 | 0.046883 | 0.046883 | 0.046883 | 1,260 | Historical exact Hero control | 0.998855 |
10,000 | 260,726 | 960 | 2,537 | 0.091019 | 0.091019 | 0.091019 | 852 | Local Search | 1 | 2,621 | -0.044137 | -0.053791 | -0.034522 | -0.044137 | -0.001145 | -0.002647 | 0 | 0.046883 | 0.046883 | 0.046883 | 725 | Historical exact Hero control | 0.998855 |
10,000 | 260,726 | 1,085 | 2,537 | 0.040134 | 0.040134 | 0.040134 | 805 | Ours (Hero) | 0.994659 | 2,621 | 0.006749 | 0.001865 | 0.011852 | 0.006749 | 0.004197 | 0.001195 | 0.007339 | 0.046883 | 0.046883 | 0.046883 | 647 | Historical exact Hero control | 0.998855 |
10,000 | 260,726 | 791 | 2,537 | 0.15078 | 0.15078 | 0.150781 | 835 | ShinkaEvolve | 0.927509 | 2,621 | -0.103898 | -0.116276 | -0.092268 | -0.103898 | 0.071347 | 0.061586 | 0.081737 | 0.046883 | 0.046883 | 0.046883 | 911 | Historical exact Hero control | 0.998855 |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Neural Solver Synthesis Final Evidence v1
This repository is the immutable large-artifact companion to the public code release for "Beyond Inference-Time Search: Reinforcement Learning Synthesizes Reusable Solvers."
Public code anchor:
07a798e7d7eca736cd1ef13a15209d402d401ef6
Current public release: NeurIPS 2026 evidence snapshot v1.0.1
Public policy checkpoints: Neural Solver Synthesis collection
The v1.0.1 patch adds checkpoint pointers only. The checksummed scientific evidence in this dataset is unchanged.
Contents
- certified full and content-clean SDS summaries;
- duplicate-safe Base Best-of-64 selection audit;
- same-model and hosted adaptive-repair controls;
- end-to-end cost and break-even accounting;
- input-disjoint universal search;
- inference-time and training-time
Hypothesizesensitivity; - compile-once JSSP summaries;
- bounded TSP RL-training results, corrected trial rows, and selected programs.
The human-readable map is docs/EVIDENCE_MAP.md. Machine-readable claims,
uncertainty conventions, sample counts, and file checksums are in
docs/final_evidence_index.json and
artifacts/neurips2026/checksums.sha256.
Verification
After downloading the repository, run the validator from the matching public
Git release with this repository's artifacts/neurips2026 directory in place:
python scripts/validate_neurips2026_public_evidence.py
Limitations and correction history
- The same-model adaptive-repair result evaluates one 64-completion execution-feedback controller and does not exhaust non-RL search.
- The hosted medium-reasoning control is not token-, dollar-, latency-, or training-compute-matched to the smaller open model. Its final allocation followed an earlier truncated attempt.
- JSSP is a within-family deployment test using JSSP-trained policies.
- TSP is direct-from-base RL training without an SFT stage. Its predeclared quality/stability gate failed, and native 2-opt and OR-Tools remained stronger.
- The TSP parser correction was applied after outcomes were observed, so TSP is boundary evidence rather than blind confirmation.
- The Qwen same-model adaptive-repair selected programs are identified by frozen SHA-256 values; their historical source files were unavailable in the compact publication workspace and were not reconstructed from test outcomes.
The evidence does not establish that RL is uniquely necessary, that transfer is tuning-free, or that generated solvers dominate native solvers.
Licensing
See docs/LICENSING.md. This evidence repository does not add a blanket license
to code, models, datasets, or standard benchmark instances.
- Downloads last month
- 42