Problems
Every problem posed by Paul Erdős, with its status, references, discussion and proof claims.
Every problem posed by Paul Erdős, with its status, references, discussion and proof claims.
110 of 1,221 problems match
Every problem posed by Paul Erdős, with its status, references, discussion and proof claims.
Asks whether the largest such that every ordering pattern of consecutive values of occurs below is , and whether the decreasing pattern always fails first.
Asks whether , where counts the totient values up to , and whether has an asymptotic formula. The doubling limit is proved twice in Lean, and an asymptotic equivalent from the OpenAI release is a pending claim.
Asks three questions comparing the least prime with the least such that . A September 2026 forum claim answers them no, no and yes, pending review.
Asks whether, for any and sets , the integers with for all with must have a logarithmic density.
Asks whether, for a finite set with set of multiples , for all ; a counterexample claim of September 2026 is pending and unreviewed.
Asks whether, for with and the integers divisible by no member of , the limit of exists and is finite.
Asks for the least number of circles determined by points in the plane, not all on one circle or one line (Elliott's reading). Purdy and Smith's correction settles every , and a 2026 claim giving every value is pending.
Asks whether every transcendental entire function has a path to infinity on which for all , how long it must be, and whether can outgrow a fixed function of its maximum modulus.
Asks whether every two-coloring of the edges of has monochromatic paths, all of one color, that cover all vertices. It is proved for , and the remaining are claimed in an unrefereed 2026 preprint.
Asks whether the number of roots in the closed unit disc of a random degree- polynomial with independent uniform coefficients satisfies almost surely.
Asks for the order of magnitude, for almost all , of , where are the binary digits of .
Asks for estimates of , the size of the largest with no distinct such that , and in particular whether .
Asks for the best possible upper bound on for in which each has at most representations with prime and .
Asks for a proof that, for large , every tree on vertices and the complete multipartite graph with class sizes satisfy .
Asks for the least constant such that for every graph with edges and no isolated vertices.
Asks for a characterization of the finite -uniform hypergraphs that appear in every -uniform hypergraph of uncountable chromatic number.
Asks whether the two Erdős–Pach–Pollack–Tuza diameter bounds hold for connected graphs with no (part (i)) or no (part (ii)). Part (i) is refuted in a refereed paper; part (ii) is proved at , and pending claims refute it for some larger .
Asks how large a triangle-free induced subgraph every -free graph on vertices must contain (the Erdős–Rogers problem). The refereed record gives the answer up to a logarithmic factor, and a 2026 preprint claims it is of order .
Asks whether every set mapping on a set of size , a function from the finite subsets of to with , has an infinite independent set.
Asks whether a family of finite graphs closed under subgraphs, with members forcing a monochromatic triangle under each finite number of colors, must contain all finite subgraphs of some graph forcing one under colors, for every infinite cardinal . A disproof is claimed.
Asks to estimate how many edges force a -uniform hypergraph on vertices to have four edges with and .
Asks whether points in the plane determine at least distinct distances when no circle centered at one of them contains three others.
Asks whether points at mutual distance at least have at most distances at most , where is the triangular lattice's count. As worded the question is garbled and fails under every counting reading, and a disproof is claimed.
Asks whether some nodes in make the Lagrange interpolants of every continuous converge to at some where the Lebesgue function is unbounded, and whether this can happen when it is unbounded at every .
Asks whether, for large , one residue class can be chosen for each prime so that every integer in lies in at least two of them. It was open on erdosproblems.com, with three pending AI-assisted full claims from April and July 2026.