Must-Read Papers of the Week: MoE Scaling Laws, World Model Physics Benchmarks, VLA Training
TheTuringPost · x · 2026-09-09
The Turing Post's weekly must-read paper roundup spans:
- Architecture & scaling: SMELT (compute-matched scaling laws for MoE looped transformers), Modern Transformers Are Implicit Hybrids
- Reasoning & training: RISE (recursive improvement via self-extrapolating policy distillation), extremely sparse supervision incentivizing reasoning, rethinking on-policy distillation of LLMs II
- World models & embodiment: World-Coherent Decoding (self-verifying test-time planning for world action models), VeriPhy (agentic physical reasoning for world model evaluation), WISE (world-model-guided imagination scheduling for VLA post-training), TourPhysics, Principia (relational physics tests for video models)
- Applications: WeatherNext 3 improving global weather model resolution from raw observations
World-model and physical-reasoning papers dominate this week's list.
More from Research
- Tabular foundation models over LLMs for predictions: talk at AI Engineer Paris — helloiamleonie · 2026-09-09
- Composing Continual Learning Mechanisms Boosts 100-Task Memorization Retention 28x to 34.9% — DanielKhashabi · 2026-09-09
- How OpenAI found a singularity in Navier-Stokes — and what it means for AI in science — eigensteve · 2026-09-09
- OpenAI's millennium proof dispute: fraud accusations, Altman denial, Tao's open science warning — The Decoder · 2026-09-09
- François Fleuret: we know training works, but not why — inductive bias, distillation and more remain opaque — francoisfleuret · 2026-09-09
- OpenWAM Releases Open Modular World-Action Model with Strong Real-Robot Performance — OpenWAM · 2026-09-09