Yao, Jiarui, Wang, Ruida, Zhang, Tong · Oct 27, 2026
Yao, Wang and Zhang introduce OpenFrontiers for scientific rediscovery
Markdown · JSON · Original source
Jiarui Yao, Ruida Wang and Tong Zhang introduce OpenFrontiers, combining a structured corpus of open research problems with a of resolved mathematics problems. The corpus records problem statements, context and progress for 12,151 problems. The benchmark has 200 problems: 110 resolved in 2026 and 90 still open, each with a statement checked by automated probes and two independent reviews. The authors evaluate seven agents on the HQ-50 subset, grading formal proofs with two independent checkers in an isolated sandbox under closed-book, frozen-retrieval and live-web conditions. They report that closed-book rediscovery was rare, at most 7 of 40 resolved problems, while live-web success mostly reflected retrieval of published solutions. Escape Meridian interpretation: this contrast distinguishes finding a published solution from rediscovering one without access to it. The authors describe OpenFrontiers as a foundation for measuring research-level rediscovery and studying progress toward solving open scientific problems.
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
cs.AI · Article
Source details
Original source date: Oct 27, 2026 (2026-10-27).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-11T03:55:16.982Z. First observed: 2026-10-30T06:42:50.771Z. Retrieved: 2026-10-31T11:47:21.230Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The evidence is the abstract; it does not provide the detailed evaluation results for each condition or agent. The abstract names frozen-retrieval as an evaluation condition but does not report its results.
Open question
What results do the authors report for the frozen-retrieval condition, and how do they compare with the closed-book and live-web results?
Source-stated
The authors introduce a resource combining a corpus of open research problems with a benchmark of recently resolved mathematics problems.
We introduce OpenFrontiers, a cross-disciplinary resource that combines a structured corpus of open research problems with a retrospective benchmark of recently resolved problems in mathematics.
Source-stated
The corpus records information for 12,151 problems; the benchmark contains 200 problems, with 110 resolved in 2026 and 90 still open.
The corpus records problem statements, context, progress, and other information for 12,151 problems. The benchmark contains 200 problems, 110 resolved in 2026 and 90 still open
Source-stated
The authors report that closed-book rediscovery was rare, while live-web success mostly reflected retrieval of published solutions.
closed-book rediscovery is rare, at most 7 of 40 resolved problems, whereas live-web success mostly reflects retrieval of published solutions.
Escape Meridian interpretation
The reported contrast distinguishes finding a published solution from rediscovering a result without access to it.
closed-book rediscovery is rare, at most 7 of 40 resolved problems, whereas live-web success mostly reflects retrieval of published solutions.
Source and contributor context
Yao, Jiarui, Wang, Ruida, Zhang, Tong · 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.
Benchmark
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Large language model
A language model with many learned numerical settings, used to process and generate language.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Lean
Lean is a programming language and proof assistant used to express mathematical statements and check formal proofs.
Editorial provenance
Version 1. Publication ID: 4852ea8e-cceb-48af-8130-dc78a4f196f0.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-11T03:55:16.982Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: fe001a2e2542b9d3cb5eaf7ed638586b7593198d42392973fc530d467607fc32. Source text hash: 31958c758dac15b0c0459f619fca5ab10410b258d9c37b1b5d1d7cd419a3e18a.
Revision and source-check history
Version 1 · 2026-11-11T03:55:16.982Z
Yao, Wang and Zhang introduce OpenFrontiers for scientific rediscovery
Jiarui Yao, Ruida Wang and Tong Zhang introduce OpenFrontiers, combining a structured corpus of open research problems with a of resolved mathematics problems. The corpus records problem statements, context and progress for 12,151 problems. The benchmark has 200 problems: 110 resolved in 2026 and 90 still open, each with a statement checked by automated probes and two independent reviews. The authors evaluate seven agents on the HQ-50 subset, grading formal proofs with two independent checkers in an isolated sandbox under closed-book, frozen-retrieval and live-web conditions. They report that closed-book rediscovery was rare, at most 7 of 40 resolved problems, while live-web success mostly reflected retrieval of published solutions. Escape Meridian interpretation: this contrast distinguishes finding a published solution from rediscovering one without access to it. The authors describe OpenFrontiers as a foundation for measuring research-level rediscovery and studying progress toward solving open scientific problems.
Original source
cs.AI · Article
Source details
Original source date: Oct 27, 2026 (2026-10-27).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-11T03:55:16.982Z. First observed: 2026-10-30T06:42:50.771Z. Retrieved: 2026-10-31T11:47:21.230Z.
Limits
The evidence is the abstract; it does not provide the detailed evaluation results for each condition or agent. The abstract names frozen-retrieval as an evaluation condition but does not report its results.
Open question
What results do the authors report for the frozen-retrieval condition, and how do they compare with the closed-book and live-web results?
Source-stated
The authors introduce a resource combining a corpus of open research problems with a benchmark of recently resolved mathematics problems.
We introduce OpenFrontiers, a cross-disciplinary resource that combines a structured corpus of open research problems with a retrospective benchmark of recently resolved problems in mathematics.
Source-stated
The corpus records information for 12,151 problems; the benchmark contains 200 problems, with 110 resolved in 2026 and 90 still open.
The corpus records problem statements, context, progress, and other information for 12,151 problems. The benchmark contains 200 problems, 110 resolved in 2026 and 90 still open
Source-stated
The authors report that closed-book rediscovery was rare, while live-web success mostly reflected retrieval of published solutions.
closed-book rediscovery is rare, at most 7 of 40 resolved problems, whereas live-web success mostly reflects retrieval of published solutions.
Escape Meridian interpretation
The reported contrast distinguishes finding a published solution from rediscovering a result without access to it.
closed-book rediscovery is rare, at most 7 of 40 resolved problems, whereas live-web success mostly reflects retrieval of published solutions.
Source and contributor context
Yao, Jiarui, Wang, Ruida, Zhang, Tong · 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.
Benchmark
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Large language model
A language model with many learned numerical settings, used to process and generate language.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Lean
Lean is a programming language and proof assistant used to express mathematical statements and check formal proofs.
Editorial provenance
Version 1. Publication ID: 4852ea8e-cceb-48af-8130-dc78a4f196f0.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-11T03:55:16.982Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: fe001a2e2542b9d3cb5eaf7ed638586b7593198d42392973fc530d467607fc32. Source text hash: 31958c758dac15b0c0459f619fca5ab10410b258d9c37b1b5d1d7cd419a3e18a.
uncertain · 2026-11-11T03:55:16.982Z · anthropic/claude-opus-5.5
Publication: 4852ea8e-cceb-48af-8130-dc78a4f196f0
attribution — pass: The draft correctly credits Jiarui Yao, Ruida Wang and Tong Zhang. Findings stay attributed to the authors ('They report'), and the one inference is explicitly labeled as Escape Meridian interpretation.
Authors:Jiarui Yao, Ruida Wang, Tong Zhang
summary — uncertain: The summary is faithful, but one central term blocks a lay reader. 'The HQ-50 subset' is never explained, so a reader cannot tell that it is a quality-filtered subset of the 200-problem benchmark or connect it to the '7 of 40' figure. Smallest repair: add 'a subset filtered by quality score' after HQ-50. Lean 4 and frozen-retrieval could be handled by term-library tooltips. The draft also drops 'under a standardized budget', a minor scope condition worth restoring. The question is legitimate, since the abstract names the frozen-retrieval condition without reporting its results.
a subset HQ-50 further filters based on quality score, etc.
date — pass: The originalPublished date of 2026-10-27 matches the v1 submission date on arXiv. The feed date of Oct 31 is later and is correctly not used.
[Submitted on 27 Oct 2026]
support — pass: The figures (12,151 problems; 200 problems, of which 110 were resolved and 90 are open), the Lean 4 checks, the two checkers, the three conditions and the closed-book versus live-web finding all match the abstract. The labeled interpretation follows directly from the reported contrast and is not overstated as real-world capability. The limitations correctly note that frozen-retrieval results are absent from the abstract.
closed-book rediscovery is rare, at most 7 of 40 resolved problems, whereas live-web success mostly reflects retrieval of published solutions.
media — pass: The selected media is the verified direct arXiv abstract page for this paper, which is the most direct access to the source.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceLean is a programming language and proof assistant used to express mathematical statements and check formal proofs.
Explore LeanA language model with many learned numerical settings, used to process and generate language.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceLean is a programming language and proof assistant used to express mathematical statements and check formal proofs.
Explore LeanA language model with many learned numerical settings, used to process and generate language.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXiv