Cross-domain data: time horizon levels differ but doubling stays at 4-7 months
BasilHalperin · x · 2026-11-09
BasilHalperin cites Kwa's (METR) important cross-domain analysis to sharpen the puzzle: METR's time horizon is measured on software tasks, but across other domains (MATH, GPQA, Tesla FSD, OSWorld, etc.) the level of the horizon differs while the speed of increase stays similar — about 4-7 month doubling.
This makes OpenAI's internal 2-month doubling even more puzzling: whether it's domain difference, stronger internal models, or methodology remains unresolved.
More from Models
- Artificial Analysis launches Decision Index v2 with 45% private data and calibration-based scoring — ArtificialAnlys · 2026-11-09
- GLM-4.5 Hit #1 on Hugging Face Within 48 Hours of MIT-Licensed Release — ccerrato147 · 2026-11-09
- Has any LLM release ever been decidedly worse than its predecessor? — adonis_singh · 2026-11-09
- CEOs say Google's Argon stands out for writing, calling Google the AI dark horse — danshipper · 2026-11-09
- Epoch AI adds ECI scores, projects Gemini 3 Pro time horizon at 4.9 hours — CFGeek · 2026-11-09
- Infosec researcher maps 3 low-hanging abuse paths for open-weight models — matthew_d_green · 2026-11-09