Computer Science > Machine Learning
[Submitted on 23 Oct 2026]
Title:RadonBench: A Benchmark of Classical and Learning-Guided Search Techniques for Problems in Combinatorial Geometry
View PDF HTML (experimental)Abstract:Classical optimization and learning-guided search offer complementary approaches to open problems in discrete mathematics, but their outputs are difficult to compare when search spaces, validity checks, and stopping rules differ. We present \emph{RadonBench}, a benchmark for constructing maximum-cardinality Radon-free lattice configurations: finite subsets of $\mathbb{Z}^d$ admitting no integral Radon partition. It remains open whether Radon-free configurations with $11$ points in $\mathbb{Z}^3$ or $21$ points in $\mathbb{Z}^4$ are possible. If they exist, they would improve the corresponding lower bounds on the integer Radon numbers. We compare exhaustive extension, greedy local search, PatternBoost, AlphaZero, and AlphaEvolve. All outputs use one success predicate: a convex target-sized configuration succeeds exactly when its number of integral Radon points is zero. PatternBoost and AlphaZero construct verified Radon-free configurations of sizes $10$ and $20$, one point short of the respective targets; AlphaZero's best convex target terminals have two integral Radon points in both dimensions. Exhaustive extension over fixed seeds and declared finite search boxes likewise reaches a minimum of two integral Radon points. Greedy local search and AlphaEvolve each produce target-sized near misses with one integral Radon point. No audited run reaches zero. These results establish reproducible near-miss baselines over explicitly scoped search domains.
Ancillary-file links:
Ancillary files (details):
- supplementary_bundle/MANIFEST.tsv
- supplementary_bundle/README.md
- supplementary_bundle/code/alphazero_3d/MCTS.py
- supplementary_bundle/code/alphazero_3d/auto_pipeline.py
- supplementary_bundle/code/alphazero_3d/net.py
- supplementary_bundle/code/alphazero_3d/radon_utils.py
- supplementary_bundle/code/alphazero_3d/set_transformer.py
- supplementary_bundle/code/alphazero_3d/set_transformer_numeric.py
- supplementary_bundle/code/alphazero_3d/set_transformer_numeric_pv.py
- supplementary_bundle/code/alphazero_3d/trainer.py
- supplementary_bundle/code/alphazero_4d/MCTS.py
- supplementary_bundle/code/alphazero_4d/auto_pipeline.py
- supplementary_bundle/code/alphazero_4d/net.py
- supplementary_bundle/code/alphazero_4d/radon_utils.py
- supplementary_bundle/code/alphazero_4d/set_transformer.py
- supplementary_bundle/code/alphazero_4d/set_transformer_numeric.py
- supplementary_bundle/code/alphazero_4d/set_transformer_numeric_pv.py
- supplementary_bundle/code/alphazero_4d/trainer.py
- supplementary_bundle/code/patternboost/box_geometry.py
- supplementary_bundle/code/patternboost/incremental_radon_saturation.py
- supplementary_bundle/code/patternboost/makemoretokens.py
- supplementary_bundle/code/patternboost/patternboost_loop.py
- supplementary_bundle/code/patternboost/radon_utils.py
- supplementary_bundle/code/patternboost/sequence3d.py
- supplementary_bundle/data/ai_10pt_3d_configs.json
- supplementary_bundle/data/alphaevolve_style/3d/mutation_results.json
- supplementary_bundle/data/alphaevolve_style/4d/mutation_results.json
- supplementary_bundle/data/alphaevolve_style/summary.json
- supplementary_bundle/data/alphazero/3d/loss_metrics.csv
- supplementary_bundle/data/alphazero/4d/loss_metrics.csv
- supplementary_bundle/data/audit/reported_results_audit.json
- supplementary_bundle/data/exhaustive_search/3d/results.json
- supplementary_bundle/data/exhaustive_search/4d/results.json
- supplementary_bundle/data/local_search/results_0_999.csv
- supplementary_bundle/data/local_search/summary.json
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen0_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen10_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen11_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen12_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen13_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen14_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen15_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen16_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen17_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen18_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen19_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen1_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen20_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen2_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen3_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen4_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen5_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen6_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen7_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen8_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_sets/B20x20x5_gen9_unpermuted.npy
- supplementary_bundle/data/patternboost/3d/training_summary.csv
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen0_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen10_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen11_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen12_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen13_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen14_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen15_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen16_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen17_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen18_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen19_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen1_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen20_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen2_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen3_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen4_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen5_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen6_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen7_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen8_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_sets/B6x6x6x6_gen9_unpermuted.npy
- supplementary_bundle/data/patternboost/4d/training_summary.csv
- supplementary_bundle/figures/ai_10pt_3d_configs.pdf
- supplementary_bundle/figures/az_training.pdf
- supplementary_bundle/figures/fixed_target_search_outcomes.pdf
- supplementary_bundle/figures/learning_guided_construction_progress.pdf
- supplementary_bundle/plotting/gen_fig_ai_10pt_3d_configs.py
- supplementary_bundle/plotting/gen_fig_az_3d_4d_training.py
- supplementary_bundle/plotting/gen_fig_fixed_target_outcomes.py
- supplementary_bundle/plotting/gen_fig_learning_guided_construction_progress.py
- supplementary_bundle/plotting/paper_plot_style.py
- supplementary_bundle/requirements.txt
- supplementary_bundle/run_metadata.json
Loading...
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
ScienceCast (What is ScienceCast?)
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
(What is IArxiv?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.