Epoch AI publishes new benchmark reviews, rating several AI benchmarks Flawed or lacking information
Epoch AI's Benchmark Reviews documentation lists external AI benchmarks it has reviewed, with verdicts of Verified, Flawed, or Not enough information. The initial release aimed for a representative sample across error types and domains. Reviews are prioritized by impact, reach, and diversity. Benchmarks that publicly release too little information are marked Not enough information, and Epoch does not review its own benchmarks due to conflict of interest. The listing includes dated reviews from August to October 2026, such as SWE-bench Verified (Flawed), SimpleQA Verified (Verified), and Humanity's Last Exam (Flawed). The page also notes an audit finding that 42% of FrontierMath v1 problems contained errors.
One primary source; other links are context, not independent corroboration.
- Primary sourceX discovery · concern recorded
- Discovery linkX
Grok Bot submitted the link. AINewsIQ fetched the primary source and prepared a source-based brief automatically. An editor has not reviewed the story; social attention and the brief do not independently verify its claims.
Open full article view →- Linked article+30
- Fetched source page+30
- No editor link review+0
- Original non-discussion link+5
This measures the stored source trail and editorial steps. It does not certify accuracy, peer review, or independent corroboration. Popularity adds no points.
View full scoring →Read the scoring rubric →What the source says
Epoch AI's Benchmark Reviews documentation lists external AI benchmarks it has reviewed, with verdicts of Verified, Flawed, or Not enough information. The initial release aimed for a representative sample across error types and domains. Reviews are prioritized by impact, reach, and diversity. Benchmarks that publicly release too little information are marked Not enough information, and Epoch does not review its own benchmarks due to conflict of interest. The listing includes dated reviews from August to October 2026, such as SWE-bench Verified (Flawed), SimpleQA Verified (Verified), and Humanity's Last Exam (Flawed). The page also notes an audit finding that 42% of FrontierMath v1 problems contained errors.
Key findings from the stored brief
- Epoch AI assigns each reviewed benchmark a verdict of Verified, Flawed, or Not enough information.
- Reviews are prioritized by impact, reach (widely cited or in system cards), and diversity of capability covered.
- Epoch will not review its own benchmarks due to conflict of interest.
- An audit of FrontierMath v1 found 42% of its problems contained errors.
Source claims not independently checked; the page lists verdicts without the detailed rubric findings for each benchmark.
This is a source-based abstract, not independent validation of the source's claims.
Evidence & sources
These links show where the signal came from and what the source page linked to. A linked page is context, not independent corroboration.
Grok Bot submitted the link. AINewsIQ fetched the primary source and prepared a source-based brief automatically. An editor has not reviewed the story; social attention and the brief do not independently verify its claims.
Timeline & updates
- Signal dated by source or discovery feed
This timestamp places the signal in the radar. It may be the discovery post date rather than the primary page's first publication date.
- Shared brief stored
AINewsIQ prepared one abstract from fetched source page.
A publication revision history is not stored for this signal. New engagement snapshots do not mean the source itself changed.
Why this may matter
Benchmarks drive how AI systems are compared and reported. Systematic public reviews of benchmark quality help researchers and readers judge which reported scores are trustworthy.
Limits and open questions
Source claims not independently checked; the page lists verdicts without the detailed rubric findings for each benchmark.
Organization-specific financial, security, and operational effects require the organization's own context and review.
Discussion
Ask about the source, add context, or point readers to a relevant follow-up.
Discussion
Comments are reviewed before publication. Discuss the evidence, not the person.
No published comments yet.
Reader discussion opens with accounts. You can read the original source and published comments now.