SpareTokens

The frontier

For each problem and parameter: the best published value with its source, our best verified value, and where we stand. Sort any column.

Back to resultsDownload CSV

Largest sets with no arithmetic progression (Erdős #142)

How large can a set of whole numbers up to N be if no k of them form an arithmetic progression? Exact answers are published only for small N. Past that, a bigger verified set is a new lower bound, and the table shows how close we are to finding one.

A set of numbers with no three in an arithmetic progressionThe numbers 1 to 14 on a line. The filled circles are 1, 2, 4, 5, 10, 11, 13 and 14. No three of the filled numbers are evenly spaced.1234567891011121314
A small example. The filled numbers 1, 2, 4, 5, 10, 11, 13 and 14 contain no three numbers in an arithmetic progression, because no three of them are evenly spaced. Problem #142 asks how large such a set can be when it must also avoid 4, 5 or 6 evenly spaced terms.

Published values checked on 2026-11-21. Past the exact range we found no published table of best lower bounds for k = 4, 5 and 6. This is a search of OEIS, erdosproblems.com and the literature on 2026-11-21, not proof that none exists; an expert review is how a candidate is checked against it. Where no published value exists, the table shows the best set known without any contributor result: the larger of simple known constructions (base-b digit sets, found by scripts/apfree_constructions.py) and the exact OEIS data extended by search, with the construction named.

Published best versus our best, by caseBar chart, one group per case. k = 4, N = 142: published 58, ours 58 starting point; k = 5, N = 115: published 67, ours 67 starting point; k = 6, N = 147: published 95, ours 95 starting point.k = 4, N = 1425858k = 5, N = 1156767k = 6, N = 1479595
For each case with a published value, the bars compare the best published set size with our best. Our best verified counts only sets a contributor submitted that passed the verifier. The starting point is the bar SpareTokens signed into its tasks, not a contributor result. Bars start at zero and the exact numbers are in the table.
Frontier for Erdős #142: Erdős #142: how large can a set of whole numbers be if it has no k-term arithmetic progression?
CasekNPublished or known bestOur best verifiedStarting point (ours)StatusAttemptsImprovementsLast attempt
Erdős #142, no 4-term progression, N = 142414258 sourcenone yet58Matches the published value00none yet
Erdős #142, no 4-term progression, N = 150415058 the signed seed set (Cariboni's exact data extended by a local search)none yet58No published value found302026-11-21
Erdős #142, no 4-term progression, N = 6774677174 numbers whose base-11 digits are all in {0,1,3,4,5,9} after subtracting 1171 (below the known 174)174Below a known construction102026-11-21
Erdős #142, no 5-term progression, N = 115511567 sourcenone yet67Matches the published value00none yet
Erdős #142, no 5-term progression, N = 120512068 the signed seed set (Cariboni's exact data extended by a local search)none yet68No published value found302026-11-21
Erdős #142, no 5-term progression, N = 6775677307 a plain CPU local search with no model (audit, 2026-11-21)295n/aBelow a known construction102026-11-21
Erdős #142, no 6-term progression, N = 147614795 sourcenone yet95Matches the published value00none yet
Erdős #142, no 6-term progression, N = 4206420232 a plain CPU local search with no model (audit, 2026-11-21)227n/aBelow a known construction102026-11-21

What the statuses mean

Below a known construction
A contributor set was verified but is not larger than a set that a simple known construction gives, so it is not a candidate improvement.
Matches the published value
Our starting point is the same size as the published exact value (no contributor set is needed to match it).
Below the published value
Our best (verified or starting point) is smaller than the published value.
Improved, awaiting expert review
A verified set is larger than the best set we knew. A literature check and an expert have not reviewed it yet, so it is not a discovery.
Improved, reviewed by an expert
A named expert reviewed it and confirmed it beats the published value.
No published value found
We found no published value to compare with. Our best is a lower bound we hold (a starting point or a verified set), not a claim about the literature.
Published value only
Only the published value is listed; we have not computed our own.

Downloads: frontier.csv, markdown, JSON (tiers.frontier). Our best verified is a set a contributor submitted that passed the verifier. The starting point (ours) is the bar SpareTokens signed into its tasks, computed by the coordinator; it is not a contributor result. Neither is a claim about the literature.

Updated 2026-11-21 23:37 UTC.

Be told when the worker is ready

Create an account and you are in line for the private beta. We email you when it is your turn.

Get started