Tasks easy to benchmark are easy to train on: if you can measure it, you can hill-climb it
gleech · x · 2026-11-11
Adam Scholl observes that tasks easy to make benchmarks about share a lot in common with tasks easy to train on: if you can measure performance on a task numerically, you can probably also train to hill-climb that metric — explaining why benchmark scores often diverge from real capability.
More from AGI Musings
- Humans will probably always set AI's highest-level goals, however smart AI gets — burny_tech · 2026-11-11
- AI agents find unmapped Maya site in Mexico using NASA LiDAR data in 72 hours — stealthispost · 2026-11-11
- TheZi slams Anthropic's "honeypot" eval as Claude Mythos 5.2 transparently answers 84 over truer 88 — TheZvi · 2026-11-11
- Designers' 'before AI' case studies will vanish from portfolios within months — round · 2026-11-11
- Railroad history warning: almost none of the first railroad companies made money — tszzl · 2026-11-11
- Fable 5.2 says the best 10-year investment is an AI-readable personal record of your thinking — churchkey · 2026-11-11