Nathan Lambert lays out six open problems in post-training in new Cornell slide deck
natolambert · x · 2026-11-21
Nathan Lambert, Executive Director of Trillium Labs, released a 51-page slide deck from a recent Cornell visit mapping the biggest open problems in post-training and how the field evolved, closing with a debate on how RSI is changing this research.
His six research areas:
- Harness specialization: designing environments for context management that enable new LLM usage paradigms
- Character training: how model personality is shaped and its impact on safety, trust, and diffusion
- Open-source RL environments: scalable, open data methods
- Reasoning efficiency in RL: controlling and manipulating reasoning efficiency
- Scaling laws for RL: principled experimentation methodology for large-scale post-training
- Multi-agent scaling laws: the main axes of inquiry into multi-LLM systems
He invites anyone working on these openly to get in touch; a re-recorded video is coming.
Related event: Nathan Lambert shares 51-page slides on open post-training problems(2 posts)→
More from AGI Musings
- AI commentator: human knowledge institutions were already rotten before AI arrived — teortaxesTex · 2026-11-21
- Fake AI breakthroughs debate: critics say hyped claims are freezing academic research — teortaxesTex · 2026-11-21
- Fields Medal reportedly denied over AI-assisted Riemann proof, sparks debate — GiorgioPatrini · 2026-11-21
- Gary Marcus: Tech Bros Lost Half Their Portfolios and Cling to Riemann Result as a Life Raft — GaryMarcus · 2026-11-21
- Mathematician blames gatekeeping culture, not candidates, in AI-math election drama — IAmTimNguyen · 2026-11-21
- Robert Wright on why AI governance shouldn't wait for its 'World War II moment' — SpencrGreenberg · 2026-11-21