The Herzog-Schönheim conjecture asserts that if a group is partitioned into finitely many left cosets
of proper subgroups of finite subgroup
index, with
,
then the indices
cannot all be distinct (Herzog and Schönheim
1974, Sun 2004). The cosets must cover the group
exactly once.
Menon (2026, Theorem 1.2) reports a proof for every solvable group, including infinite groups. The author credits AI with developing parts of the proof and completing a Lean formalization.
The conjecture remains open for arbitrary groups (Bloom 2026). For infinite groups, equality of two indices concerns the numbers of cosets, rather than the possibly infinite sizes of the subgroups.
This is problem 274 in the Erdős problems collection (Bloom 2026).