# The FrontierOR Challenge: Agentic Algorithm Design for Operations Research

[Canonical reading](https://escapemeridian.com/readings/the-frontieror-challenge-agentic-algorithm-design-for-3e6ed88e)

The authors of “The FrontierOR Challenge: Agentic Algorithm Design for Operations Research” argue that whether large language model (LLM) agents can design algorithms that scale to realistic operations research problems remains an open question. They cite a one-shot evaluation in which the strongest model matched Gurobi—within 1% of its objective and no slower—on only 39% of large instances. The challenge is designed to test whether agentic workflows can design, test and improve algorithms across many optimization problems. Its first stage releases part of an audited benchmark in weekly rounds; a second stage uses new problems outside that benchmark and unseen instances. The challenge separates a Main track, where submitted algorithms run in a network-isolated sandbox with open-source dependencies, from a Self-hosted track, where teams solve instances on their own machines using any software within a timed window. The abstract describes the challenge design, not results showing how well participants’ algorithms perform.

Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.

## Original source

arXiv cs.AI · Article

Original source date: Nov 19, 2026 (2026-11-19).

Date basis: Date basis not provided.

Added to Escape Meridian: 2026-11-20T08:47:27.192Z. First observed: 2026-11-20T05:07:44.268Z. Retrieved: 2026-11-20T08:46:16.863Z.

- [The FrontierOR Challenge: Agentic Algorithm Design for Operations Research](https://arxiv.org/abs/2611.26171)

## Limits

The supplied evidence is the arXiv abstract; it does not provide challenge results or further detail about the benchmark and evaluation beyond the description quoted here.


## Open question

How will submitted workflows perform on Stage 2’s new problems and unseen instances?


## Source-stated

The authors say whether LLM agents can design algorithms that scale to realistic operations research instances is still an open question.

> but whether they can design algorithms that scale to realistic instances is open

- [Supporting source](https://arxiv.org/abs/2611.26171)

## Source-stated

In the cited one-shot evaluation, the strongest model met the stated comparison with Gurobi on 39% of large instances.

> the strongest model matches Gurobi, with an objective within 1% of Gurobi's and no slower, on only 39% of the large instances.

- [Supporting source](https://arxiv.org/abs/2611.26171)

## Source-stated

The challenge’s second stage introduces problems outside the benchmark, including unseen instances.

> Stage 2 introduces new problems outside the benchmark, with unseen instances.

- [Supporting source](https://arxiv.org/abs/2611.26171)

## Source-stated

The authors describe separate Main and Self-hosted tracks with different execution conditions.

> in the Main track we run the submitted algorithms in a network-isolated sandbox with open-source dependencies only; in the Self-hosted track teams solve the instances on their own machines, with any software, within a timed window.

- [Supporting source](https://arxiv.org/abs/2611.26171)

## Source and contributor context

Wang, Hai, Hajek, Bretislav, Jiang, Chonghe, Kong, Minwei, Qu, Ao, Gong, Huatian, Dong, Bo, Barnhart, Cynthia, Zhao, Jinhua · Article author

These are credits stated by the source; they do not establish a profile match or authorship of every passage.


## Source media

No suitable media was found by the supported page extraction.


## Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.


## Agentic AI

AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.

- [Definition reference](https://developers.google.com/machine-learning/glossary#agentic)

## Large language model

A language model with many learned numerical settings, used to process and generate language.

- [Definition reference](https://developers.google.com/machine-learning/glossary#large-language-model)

## Benchmark

A defined set of tests used to compare systems. Results depend on the tasks and measures selected.

- [Definition reference](https://arxiv.org/abs/2211.09110)

## arXiv

arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.

- [Explore this context](/entities/arxiv)
- [Reference](https://arxiv.org/)
- [Reference](https://info.arxiv.org/help/api/index.html)

## Editorial provenance

Version 1. Publication ID: 2d02b4e7-b043-4677-83c9-5023852af69e.

Prepared by gpt-6-luna; stream-sentence-removal-v1: removed summary 6; grounding review by gpt-6-luna.

Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-20T08:47:27.192Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: material\_development.

Observation: 505ff0fe353b3b80fb1eb3e6596f12c311e76d54642ff09e3a7e63c07b6d555f. Source text hash: c97a92ec4bc9fc5c5bec8565aa76cdd2b8fe4d8fde1fbdd1f849ceec7fa84fde.


## Revision and source-check history

### Version 1 — 2026-11-20T08:47:27.192Z

The authors of “The FrontierOR Challenge: Agentic Algorithm Design for Operations Research” argue that whether large language model (LLM) agents can design algorithms that scale to realistic operations research problems remains an open question. They cite a one-shot evaluation in which the strongest model matched Gurobi—within 1% of its objective and no slower—on only 39% of large instances. The challenge is designed to test whether agentic workflows can design, test and improve algorithms across many optimization problems. Its first stage releases part of an audited benchmark in weekly rounds; a second stage uses new problems outside that benchmark and unseen instances. The challenge separates a Main track, where submitted algorithms run in a network-isolated sandbox with open-source dependencies, from a Self-hosted track, where teams solve instances on their own machines using any software within a timed window. The abstract describes the challenge design, not results showing how well participants’ algorithms perform.

## Original source

arXiv cs.AI · Article

Original source date: Nov 19, 2026 (2026-11-19).

Date basis: Date basis not provided.

Added to Escape Meridian: 2026-11-20T08:47:27.192Z. First observed: 2026-11-20T05:07:44.268Z. Retrieved: 2026-11-20T08:46:16.863Z.

- [The FrontierOR Challenge: Agentic Algorithm Design for Operations Research](https://arxiv.org/abs/2611.26171)

## Limits

The supplied evidence is the arXiv abstract; it does not provide challenge results or further detail about the benchmark and evaluation beyond the description quoted here.


## Open question

How will submitted workflows perform on Stage 2’s new problems and unseen instances?


## Source-stated

The authors say whether LLM agents can design algorithms that scale to realistic operations research instances is still an open question.

> but whether they can design algorithms that scale to realistic instances is open

- [Supporting source](https://arxiv.org/abs/2611.26171)

## Source-stated

In the cited one-shot evaluation, the strongest model met the stated comparison with Gurobi on 39% of large instances.

> the strongest model matches Gurobi, with an objective within 1% of Gurobi's and no slower, on only 39% of the large instances.

- [Supporting source](https://arxiv.org/abs/2611.26171)

## Source-stated

The challenge’s second stage introduces problems outside the benchmark, including unseen instances.

> Stage 2 introduces new problems outside the benchmark, with unseen instances.

- [Supporting source](https://arxiv.org/abs/2611.26171)

## Source-stated

The authors describe separate Main and Self-hosted tracks with different execution conditions.

> in the Main track we run the submitted algorithms in a network-isolated sandbox with open-source dependencies only; in the Self-hosted track teams solve the instances on their own machines, with any software, within a timed window.

- [Supporting source](https://arxiv.org/abs/2611.26171)

## Source and contributor context

Wang, Hai, Hajek, Bretislav, Jiang, Chonghe, Kong, Minwei, Qu, Ao, Gong, Huatian, Dong, Bo, Barnhart, Cynthia, Zhao, Jinhua · Article author

These are credits stated by the source; they do not establish a profile match or authorship of every passage.


## Source media

No suitable media was found by the supported page extraction.


## Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.


## Agentic AI

AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.

- [Definition reference](https://developers.google.com/machine-learning/glossary#agentic)

## Large language model

A language model with many learned numerical settings, used to process and generate language.

- [Definition reference](https://developers.google.com/machine-learning/glossary#large-language-model)

## Benchmark

A defined set of tests used to compare systems. Results depend on the tasks and measures selected.

- [Definition reference](https://arxiv.org/abs/2211.09110)

## arXiv

arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.

- [Explore this context](/entities/arxiv)
- [Reference](https://arxiv.org/)
- [Reference](https://info.arxiv.org/help/api/index.html)

## Editorial provenance

Version 1. Publication ID: 2d02b4e7-b043-4677-83c9-5023852af69e.

Prepared by gpt-6-luna; stream-sentence-removal-v1: removed summary 6; grounding review by gpt-6-luna.

Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-20T08:47:27.192Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: material\_development.

Observation: 505ff0fe353b3b80fb1eb3e6596f12c311e76d54642ff09e3a7e63c07b6d555f. Source text hash: c97a92ec4bc9fc5c5bec8565aa76cdd2b8fe4d8fde1fbdd1f849ceec7fa84fde.


### uncertain — 2026-11-20T08:47:27.192Z — anthropic/claude-opus-5.5

Publication: 2d02b4e7-b043-4677-83c9-5023852af69e

attribution — pass: The reading correctly attributes the argument and challenge description to the paper's authors. It presents the 39% figure as the authors' cited one-shot evaluation, not as an independent fact.

> Authors:Hai Wang, Bretislav Hajek, Chonghe Jiang, Minwei Kong, Ao Qu, Huatian Gong, Bo Dong, Cynthia Barnhart, Jinhua Zhao

summary — uncertain: Gurobi is central to the key statistic but is never explained; a general reader cannot tell that it is the comparison reference. Smallest repair: add a tooltip or term-library entry for Gurobi. The reading also drops the source's human stake: OR algorithms are expert-tailored, and a new problem can take weeks of specialist work. Without that, readers lose why agentic algorithm design matters. Restore it in one sentence. Also restore the scoring rule: feasibility is checked, then a feasible solution scores its objective ratio to a frozen reference.

> a new routing, scheduling or network design problem can take weeks of specialist work before it is solved well at scale

date — pass: originalPublished 2026-11-19 matches the v1 submission date. The feed date of Nov 20 is correctly not used.

> \[Submitted on 19 Nov 2026\]

support — pass: Each claim is supported by a verbatim excerpt. The note that the abstract gives design rather than results is accurate. The question about Stage 2 performance follows from the evidence without assuming an outcome.

> Stage 2 introduces new problems outside the benchmark, with unseen instances.

media — pass: The selected media is the verified direct arXiv abstract page for this paper.

## Recorded references

Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter.

No public references recorded here yet.

[Reference data (JSON)](/readings/3e6ed88e764ac620ffeab5802f08c35f16be0645a4bf950e4b68ddb342ff6690/references.json)

## Corrections and challenges

[Send a correction or an opposing source](https://escapemeridian.com/corrections/)

[JSON record](https://escapemeridian.com/readings/the-frontieror-challenge-agentic-algorithm-design-for-3e6ed88e.json)
