Cohere Labs talk: Goodfire's Matthew Kowal on interpretability across the model design pipeline
Cohere_Labs · x · 2026-10-20
Cohere Labs' Safety & Alignment community hosts a talk by Goodfire researcher Matthew Kowal on Oct 23: Interpretability Across the Model Design Pipeline.
He frames interpretability through model affordances — the actions a system makes possible. Part one covers recent methods that decompose model activations into multidimensional manifolds rather than straight lines, creating better "knobs" to understand and steer model behavior. Part two covers ongoing work using these interpretable knobs during training itself, shaping learning with a model's own representations while keeping the process interpretable and monitorable.
More from Safety
- Researcher: universal multimodal jailbreaks seem near-impossible across VLMs — RylanSchaeffer · 2026-10-20
- Watch top LLMs battle it out live in an attack/defense CTF — moyix · 2026-10-20
- EMNLP study: LLM safety interventions drop up to 4.5x when distress is wrapped in delusion — MaartenSap · 2026-10-20
- Polymarket gives just 7% odds of a US AI safety bill by end of 2026 — Polymarket · 2026-10-20
- FDA Asks How to Judge Medical AI; Yale's Krumholz: Get the Comparator Right — zakkohane · 2026-10-20
- NIST AI 600-1 explained: turning 12 generative AI risks into six actionable steps — anacondainc · 2026-10-20