Signals 4 · free daily AI digest

On the estimation and validity of AI time horizons---a statistical look at the METR plot

Paper recorded by Signals 4 on 2026-10-08 in cs.AI. Abstract reproduced from arXiv; link to the original below.

Published 2026-10-08 on arXiv · recorded by Signals 4 on 2026-10-09

Category: cs.AI · 人工智能 · first seen 2026-10-09

Abstract

METR's 50\% time horizon measures the human completion time of software tasks that an AI solves with 50\% probability, allowing AI capabilities to be expressed in interpretable units. On 228 tasks and 26 AIs, we recompute the time horizons using splines and item-response theory to relax the assumption that the AI difficulty of a task depends linearly on the log of human time. Our fitted spline can

Read on arXiv →

#361 most recent of 960 cs.AI papers we have recorded · ↑ newer: Orthros: Phase-Aware Heterogeneous Attention for Efficient Transformer · ↓ older: From Reactive Containment to Proactive Assurance: Lessons from OpenAI,

Cite this page: On the estimation and validity of AI time horizons---a statistical look at the METR plot: the #361 most recent of 960 cs.AI papers we have recorded (as of 2026-10-08). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/papers/on-the-estimation-and-validity-of-ai-time-horizons-a-statistical-look-at-the-met.html
Free to quote with attribution to “Signals 4 (Signals API)”. Machine-readable: papers.json

Related: More cs.AI papers · arXiv signals · All papers · Today in AI

Get 4 AI signals a day by email — free.

Subscribe free → See all plans →

Get 4 AI signals a day by email — free

All models · All repos · By company · Daily editions