Integrality gaps in the skiving stock problem
Given items of specified sizes, divide some of them into as many disjoint groups as possible, each reaching a fixed minimum size. This is one-dimensional bin covering, also called the skiving stock problem. A pattern linear program relaxes the requirement that each group be used an integer number of times. How far can that fractional upper bound exceed the true maximum? Zak conjectured that the gap for the ordinary pattern relaxation was always below two.
Why it matters and what counts as progress
Why it matters: Pattern relaxations give upper bounds without constructing an integer solution. The size of their gap controls how much confidence a near-optimal fractional solution gives about how many groups can actually be covered.
What counts as progress: Give a bound that holds for a stated class of item sizes, or give explicit instances with a proved fractional value and a matching or separated integer optimum. Identify whether the relaxation allows patterns that use more copies of an item size than the input supplies. State which prior results the proof assumes.
How the problem developed
- 1984 · Bin covering formulation [1]
Susan F. Assmann, David S. Johnson, Daniel J. Kleitman and Joseph Y.-T. Leung Studied the dual of one-dimensional bin packing, where each counted group has to reach a minimum total size.
- 24 Oct 2003 · Modified integer round-down conjecture [2]
E. J. Zak Proposed that the ordinary fractional pattern bound exceeds the integer optimum by less than two.
- 2016 · Divisible-size theorem [3]
John Martinovic and Guntram Scheithauer Proved the modified integer round-down property for instances with divisible item lengths.
- 24 Sept 2026 · Bin-packing input [4]
OpenAI’s internal research model Reported unbounded configuration-LP gaps and additive hardness for the distinct bin-packing problem. Winnington uses the hard instances in his later transfer.
- 7 Oct 2026 · Conditional unbounded gaps [5]
Eric Winnington with Claude Fable 5.1 and GPT-5.6 Sol Pro Reported that a complement transformation of OpenAI’s bin-packing instances gives unbounded gaps for both covering relaxations and fixed-additive-allowance hardness.
AI contribution
Unbounded Integrality Gaps and Additive Hardness for the Skiving Stock Problem
Assuming the cited OpenAI bin-packing results, for every integer there is a rational one-dimensional bin-covering instance with 5B items and integer B, for which both pattern LP relaxations have value B but at most disjoint groups reach the covering threshold. The transfer also makes every fixed additive allowance NP-hard.
Sources
- On a dual version of the one-dimensional bin packing problem
- The skiving stock problem as a counterpart of the cutting stock problem
- Integer rounding and modified integer rounding for the skiving stock problem
- Additive hardness and unbounded configuration gaps in bin packing
- Unbounded Integrality Gaps and Additive Hardness for the Skiving Stock Problem
- Winnington’s proof-generation disclosure
- Winnington’s announcement
- New theoretical investigations on the gap of the skiving stock problem
Publication history
Unbounded Integrality Gaps and Additive Hardness for the Skiving Stock Problem
Daily emails by default, only when there’s an update.