Curriculum learning

Technique · Training · 4 items since May 2026

Ordering training data or tasks from easy to hard to make learning more efficient.

Organizations: Microsoft Weibo
Techniques: Harness optimization Reinforcement learning with verifiable rewards Scaling laws Self-distillation Test-time scaling

Trends

Oct 2
Microsoft's ActiveSaddler
Formulates agent-harness optimization as automated curriculum learning, modeling the evolving training curriculum as a non-stationary bandit over reusable failure-pattern arms, and lifts test Pass@1 by 4.4 points on GAIA2 and 7.5 points on Terminal-Bench 2.0 over a fixed scenario order.
Jun 15
Weibo AI's VibeThinker-3B
A 3B dense open-weight reasoning model scoring 94.3 on AIME 2026 (97.1 with test-time scaling) and 80.2 Pass@1 on LiveCodeBench v6, matching far larger flagships such as DeepSeek V3.2 and Gemini 3 Pro via curriculum SFT, multi-domain RL, and self-distillation.
May 14
SU-01 reaches olympiad gold at 30B-A3B
A 30B-A3B MoE trained with reverse-perplexity curriculum SFT, coarse RLVR then proof-level RL, and test-time verification reaches gold on IMO 2025, USAMO 2026, and IPhO 2024–25, scoring 35/42 and matching the top human among 340 USAMO competitors.
May 6
ScaleLogic finds a power law for RL compute
RL training compute scales as a power of reasoning depth (R² > 0.99), with the exponent rising from 1.04 to 2.60 as logical expressiveness grows; easy-to-hard curricula improve efficiency and lift downstream math and reasoning benchmarks by up to 10.66 points.