RECORD
RadonBench: A Benchmark of Classical and Learning-Guided Search Techniques for Problems in Combinatorial Geometry
Machine Learning · Submitted 23 Oct 2026
abstract compiled · 2026-10-27 · no language-model review
The authors' main claim These results establish reproducible near-miss baselines over explicitly scoped search domains.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
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.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | 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. | background 0.62 | 0.06 | 0.03 | asserted 1.00 |
| s1 | 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. | contribution 1.00 | 0.87 | 0.05 | asserted 1.00 |
| s2 | It remains open whether Radon-free configurations with $11$ points in $\mathbb{Z}^3$ or $21$ points in $\mathbb{Z}^4$ are possible. | gap 0.76 | 0.07 | 0.05 | asserted 0.58 |
| s3 | If they exist, they would improve the corresponding lower bounds on the integer Radon numbers. | background 0.64 | 0.09 | 0.05 | hedged possibility 0.65 |
| s4 | We compare exhaustive extension, greedy local search, PatternBoost, AlphaZero, and AlphaEvolve. | method 0.84 | 0.89 | 0.03 | asserted 0.97 |
| s5 | All outputs use one success predicate: a convex target-sized configuration succeeds exactly when its number of integral Radon points is zero. | method 0.98 | 0.86 | 0.05 | asserted 1.00 |
| s6 | 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. | result 1.00 | 0.92 | 0.11 | asserted 1.00 |
| s7 | Exhaustive extension over fixed seeds and declared finite search boxes likewise reaches a minimum of two integral Radon points. | result 1.00 | 0.89 | 0.05 | asserted 1.00 |
| s8 | Greedy local search and AlphaEvolve each produce target-sized near misses with one integral Radon point. | result 1.00 | 0.93 | 0.11 | asserted 1.00 |
| s9 | No audited run reaches zero. | result 1.00 | 0.92 | 0.50 | asserted 1.00 |
| s10 | These results establish reproducible near-miss baselines over explicitly scoped search domains. | interpretation 0.90 | 0.88 | 0.45 | asserted 1.00 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original 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.
Human review
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.