korshunov.ai
English
Benchmark · math

AIME 2025

8 results 6 models
0 7.5 15 22.5 30 2026-06-18 2026-06-21 2026-06-24 STARE · 8 · 2026-06-18 STARE · 8 · 2026-06-18 STARE · 8 · 2026-06-18 greedy router · 19.1 · 2026-06-19 baseline · 19.6 · 2026-06-19 stochastic router · 28.1 · 2026-06-19 adaptive prompt selection mechanism · 28.1 · 2026-06-19 IW-OPD · 6.9 · 2026-06-24
STARE greedy router baseline stochastic router adaptive prompt selection mechanism IW-OPD
Timeline
  1. 2026-06-24 IW-OPD 6.9pts Importance-Weighted On-Policy Distillation Addresses Position Bias
  2. 2026-06-19 adaptive prompt selection mechanism 28.1% Adaptive LLM Tutoring Improves Engagement and Efficiency
  3. 2026-06-19 greedy router 19.1% Adaptive LLM Tutoring Improves Engagement and Efficiency
  4. 2026-06-19 baseline 19.6% Adaptive LLM Tutoring Improves Engagement and Efficiency
  5. 2026-06-19 stochastic router 28.1% Adaptive LLM Tutoring Improves Engagement and Efficiency
  6. 2026-06-18 STARE 8.0% STARE: Surprisal-Guided Token-Level Advantage Reweighting for Policy Entropy Stability
  7. 2026-06-18 STARE 8.0% STARE: Surprisal-Guided Token-Level Advantage Reweighting for Policy Entropy Stability
  8. 2026-06-18 STARE 8.0% STARE: Surprisal-Guided Token-Level Advantage Reweighting for Policy Entropy Stability