LLMs Went from Failing Simple Reasoning Tasks to Solving Them All in Two Years

Stefanos Batzoglou recalled that reasoning tasks top models could only solve 10% of two years ago are now all solved by late 2025, while Anshul Kundaje countered that frontier model debates should rely on empirical evidence rather than intuition.

2026-09-09 ~ 2026-09-09 · 3 related posts

• Researcher: Judge Frontier Models on Today's Evidence, Not Hunches About the Future — anshulkundaje · 2026-09-09
• Easy Reasoning Tasks Top Models Failed 2 Years Ago Are Now Fully Saturated — s_batzoglou · 2026-09-09
• Stanford Prof Bets LLMs Reach Navier-Stokes Level in 2 Years; Biology Researcher Pushes Back — anshulkundaje · 2026-09-09