FunSearch and Structured SAT Search for Weak Schur Partitions

The weak Schur problem asks how large an interval of consecutive integers can be partitioned into a fixed number of classes with no class containing a monochromatic solution to $a+b=c$ with $a\neq b$. We investigate whether FunSearch, an evolutionary program-search method guided by a large language model, can discover effective construction strategies for this classical combinatorial problem. In our adaptation, FunSearch evolves a priority function used by a fixed greedy solver, while a separate validity check enforces the weakly sum-free condition. This approach matches the known exact values for two, three, and four classes and reaches $WS(5)\geq195$. We then explore a complementary SAT-based approach. Guided by structural information from a 207-element construction communicated by Fred Rowley (Rowley, 2026), our SAT search produces and verifies a distinct valid five-class partition of $\{1,\ldots,207\}$. Thus the strongest construction obtained in this project reaches $WS(5)\geq207$, eleven integers beyond the best published lower bound of 196. The contrast between the two approaches also exposes a limitation of our FunSearch formulation: its greedy solver makes local, irreversible decisions, whereas SAT can enforce global structure across the whole partition.

Authors

Publication Details

Journal
arXiv (Cornell University)
Published
2026-11-17
DOI
https://doi.org/10.48550/arxiv.2611.22032
Primary Topic
Constraint Satisfaction and Optimization
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

FunSearch and Structured SAT Search for Weak Schur Partitions

Eric Ramos, Jingxuan Feng, Caleb Hartman, Pablo Christofferson et al.
arXiv (Cornell University)
Constraint Satisfaction and Optimization
preprint

FunSearch and Structured SAT Search for Weak Schur Partitions

Eric Ramos, Jingxuan Feng, Caleb Hartman, Pablo Christofferson, Omar M. Elshinawy, Suleman Khan, Rebecca Vicenzi
preprint en

Abstract

The weak Schur problem asks how large an interval of consecutive integers can be partitioned into a fixed number of classes with no class containing a monochromatic solution to $a+b=c$ with $a\neq b$. We investigate whether FunSearch, an evolutionary program-search method guided by a large language model, can discover effective construction strategies for this classical combinatorial problem. In our adaptation, FunSearch evolves a priority function used by a fixed greedy solver, while a separate validity check enforces the weakly sum-free condition. This approach matches the known exact values for two, three, and four classes and reaches $WS(5)\geq195$. We then explore a complementary SAT-based approach. Guided by structural information from a 207-element construction communicated by Fred Rowley (Rowley, 2026), our SAT search produces and verifies a distinct valid five-class partition of $\{1,\ldots,207\}$. Thus the strongest construction obtained in this project reaches $WS(5)\geq207$, eleven integers beyond the best published lower bound of 196. The contrast between the two approaches also exposes a limitation of our FunSearch formulation: its greedy solver makes local, irreversible decisions, whereas SAT can enforce global structure across the whole partition.

arXiv (Cornell University)
Constraint Satisfaction and Optimization
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon