遇见数据集

Claim Verification: "AI progress in capabilities has largely plateaued since late 2024" — Disproved

收藏
Zenodo2026-04-09 更新2026-05-26 收录
官方服务:

资源简介:

Automated fact-verification of the claim: "AI progress in capabilities has largely plateaued since late 2024" Verdict: DISPROVED Key Findings The Epoch Capabilities Index (ECI) rate of improvement nearly doubled from ~8 points/year to ~15 points/year after April 2024 — the opposite of a plateau (B1, B2). On FrontierMath (advanced math), AI scores rose from <2% (November 2024) to 47.6% (March 2026) — a >20x improvement in 16 months. On SWE-bench Verified (real-world coding), top scores reached 80.9% by late 2025 and continue climbing (B3). 4 of 4 consulted sources provide quantitative evidence of continued or accelerating AI capability improvements, exceeding the threshold of 3 needed for disproof. Files proof.py — Re-runnable Python verification script proof.md — Structured proof report proof_audit.md — Full verification audit trail proof_narrative.md — Plain-language summary proof.json — Machine-readable structured data Generated by Proof Engine v1.2.0.

提供机构:
Zenodo
创建时间:
2026-04-09
二维码
社区交流群
二维码
科研交流群
商业服务