AI Proof Watch and related services
AI Proof Watch tracks reported AI contributions to research. These services track mathematical claims or evaluate AI systems, so their coverage and methods differ.
“Closest direct alternative” means a similar contribution-tracking purpose. “Adjacent” means a different service that can help readers assess AI capability.
At a glance
On smaller screens, scroll the table horizontally to see every comparison.
| Service | Main purpose | Coverage | Assessment | Email updates |
|---|---|---|---|---|
| AI Proof WatchResearch contribution tracker | Curate reported AI research contributions and list linked human follow-up works. | Mathematics, theoretical computer science, theoretical physics. | Four separate ratings: AI involvement, technical verification, human understanding, and importance. | Weekly all-topic digest; selected-interest alerts are daily by default and sent only when there is an update. |
| Proof of ProgressClosest direct alternative | Classifies reported mathematical claims against a human mathematical frontier and charts evidence over time. | Mathematics | Claim-level categories such as Expected, Difficult, and Superhuman, with confidence and evidence fields. Its page describes the collection as AI-assisted and not an expert-reviewed census. | The page lists collection refreshes and analysis history. We found no email subscription on the reviewed page. |
| What AI FoundDirect research-contribution tracker | Help readers distinguish the strength of evidence for a result from how much the AI contributed. | Mathematics, theoretical computer science, physics, medicine, biology, and other sciences | Each finding receives two independent grades: a verification grade, from claimed to formally verified or refuted, and an autonomy grade, from search scaffold to autonomous. The registry links primary work, announcements, coverage, commentary, and challenges. | The registry shows recent additions and changes. We found no email subscription on the reviewed pages. |
| VibeMathedDirect mathematics tracker | Track AI-involved mathematical results, including their status, sources, and review history. | Mathematics, including theoretical computer science | Its scope centers on precisely stated open questions answered by a theorem, with AI substantively involved. Entries distinguish resolved, partial, candidate, and other outcomes, and include separate AI-role, verification, significance, and notability labels. | The site publishes new and revised entries and has reader submissions with review. We found no email subscription on the reviewed pages. |
| IMProofBenchResearch-math benchmark | Measures model performance on a private, regularly updated set of mathematical problems. | Mathematics | Expert mathematicians grade solutions on a 0–3 scale; a score of 3 means a complete solution. The project also describes machine-checkable subproblems. | The leaderboard says the private problem set is regularly updated. We found no email subscription on the reviewed pages. |
| FrontierMath: Open ProblemsResearch-math benchmark | Tracks whether AI systems can solve selected open problems and advance mathematical knowledge. | Mathematics, with a constructive and computationally verifiable bias | Working mathematicians propose problems and an editorial board assigns notability tiers. Custom verifiers check candidate solutions; a positive verifier result can be evidence rather than a full proof. Solved problems are retired. | Epoch tests selected frontier models and updates individual problem records. We found no email subscription on the reviewed benchmark pages. |
| METRAdjacent capability evaluation | Studies broad autonomous capabilities and AI systems’ ability to accelerate AI research and development. | Long-horizon tasks across domains; not a research-subject catalog | Its time-horizon measure estimates the duration of tasks, based on human expert completion time, that an AI agent can complete at a specified reliability. | METR publishes evaluations and research reports. We found no topic-based email update product on the reviewed pages. |
| Artificial AnalysisAdjacent model benchmark and comparison service | Helps users compare AI systems and the infrastructure that serves them. | General AI models and systems, including mathematics and scientific reasoning evaluations | Publishes benchmark results and composite indices. Its Intelligence Index aggregates evaluations across agents, coding, scientific reasoning, and general tasks. | The site publishes changing benchmark and model data. We found no research-topic email alert product on the reviewed pages. |
The comparison describes publicly documented features on 20 October 2026. “We found no email subscription” means one was not identified on the linked pages; it does not establish that none exists elsewhere.
Where AI Proof Watch fits
Use AI Proof Watch when you want to follow what AI has contributed to research problems, inspect original reports and checking evidence, and judge each contribution across distinct dimensions. Use benchmarks when you want standardized model evaluations on a selected problem set or task suite.
Where AI Proof Watch may not fit
Choose another service when your main goal is:
- Follow AI-attributed work across more scientific fields: What AI Found.
- Explore community-reported AI advances on open math problems: VibeMathed.
- Compare mathematical claims with a human frontier and follow them over time: Proof of Progress.
- Benchmark models on research-level math problems: IMProofBench or FrontierMath: Open Problems.
- Measure long-task agent capability: METR.
- Compare general model performance, speed, and cost: Artificial Analysis.
Where arXiv fits
arXiv is an open research-sharing, search, and discovery platform. It is an important source of papers for AI contribution trackers, but it does not operate a dedicated registry that rates AI’s role in research results. That makes it a research source rather than a direct competitor.
Detailed comparisons
- Proof of Progress — Closest direct alternative
- What AI Found — Direct research-contribution tracker
- VibeMathed — Direct mathematics tracker
- IMProofBench — Research-math benchmark
- FrontierMath: Open Problems — Research-math benchmark
- METR — Adjacent capability evaluation
- Artificial Analysis — Adjacent model benchmark and comparison service
Read AI Proof Watch’s review criteria or choose email updates.