Model Architecture & Training MethodsSardana, Frankle et al.

Beyond Chinchilla-Optimal

Beyond Chinchilla-Optimal

Why it matters

Included inference cost in the objective, providing the theoretical basis for today's overtraining norm

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2401.00448

Added in Week 35, 2026 · See that week

Same week

NYT sues Microsoft and OpenAIPressKorean government publishes a generative AI copyright guideOfficial

Nearby on the same axis

MambaPaperSakana AI announces Evolutionary Model Merge, releases EvoLLM-JP and EvoVLM-JPOfficialStreamingLLMPaperYaRNPaper