New Scaling Law 'Skaling' Restores Interaction Between Model Size and Data
TimDarcet · x · 2026-08-26
Chinchilla scaling law treats loss as the sum of two independent terms—model size and data—ignoring their interaction, which bends predictions when extrapolating. The new 'Skaling' law reintroduces the interaction with a single exponent to correct these predictions.
More from Research
- Context compaction destroys 53% of safety rules; Knowledge Triage fixes it — omarsar0 · 2026-08-26
- Dribbling the AI Watermark Directly In-Prompt — JulianHabekost · 2026-08-26
- Reducing Sycophancy in Qwen and Gemma using Runtime Activation Steering — ASL_Dev · 2026-08-26
- Energy-first AI hardware design might mimic the brain — prateekj · 2026-08-26
- Paper feeds now support filtering by custom date range — NielsRogge · 2026-08-26
- Weekly Recap: Qwen 4, Wan 3.0, Open On-Device TTS, and Apple M6 — fromourback · 2026-08-26