Model Behavior Hinges on Post-Training
gerardsans · x · 2026-07-19
The discussion emphasizes that behavioral differences in models stem not just from pre-training data and architecture, but also from post-training and RLHF.
The author points out that while many labs share common training corpora, the exact composition is opaque; furthermore, the data and rules used during post-training are often kept strictly behind closed doors. Therefore, evaluating model performance requires looking beyond pretraining to include post-training and the alignment process.
Related event: Claude's Constitutional AI: Post-Training Dictates Model Behavior(2 posts)→
More from Research
- MissionReport benchmark: only 3 of 540 AI-generated government reports pass all criteria — Exp_Mark · 2026-10-27
- Finch 1: a geospatial foundation model trained on imagery from 50+ sensors — Scobleizer · 2026-10-27
- Skyfall AI's SKY-001 world model turns sketches into CAD builds in 3.9s, vs 199s for GPT-6 Astra — rohanpaul_ai · 2026-10-27
- Physicists Let AI Pick Its Own Problem—It Wrote a Publishable Paper in 2 Hours — burny_tech · 2026-10-27
- Researcher flags unresolved data transparency issues in compositionality evals — najoungkim · 2026-10-27
- Separate Forms from Estimators: Why Steer by Difference of Means When You Can Use SGD? — johnhewtt · 2026-10-27