thebes on X: "i dislike how people talk about model looping in a binary way when it comes to monitorability when what actually matters is effective depth and how it affects verbalization necessity afaict the chain of thought for experienced researchers concerned about the news is something https://t.co/6iLLCTUOmd" / X
i dislike how people talk about model looping in a binary way when it comes to monitorability when what actually matters is effective depth and how it affects verbalization necessity afaict the chain of thought for experienced researchers concerned about the news is something like: "previously, h…
i dislike how people talk about model looping in a binary way when it comes to monitorability when what actually matters is effective depth and how it affects verbalization necessity afaict the chain of thought for experienced researchers concerned about the news is something like: "previously, hardware limited parameter count, and parameter count + a reasonable aspect ratio limited maximum depth. looping lets you reuse the same parameters for more depth, removing this ceiling, allowing for higher effective depth on the existing hardware and reducing the 'necessity' arm of CoT…
saved by
related reading
- [2512.18311] Monitoring Monitorabilityarxiv.org
- [2507.11473] Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safetyarxiv.org
- How AI Is Learning to Think in Secretnickandresen.substack.com
- Why We Are Excited About Confessionsalignment.openai.com
- Can activation verbalizers surface an internal chain of thought? — LessWronglesswrong.com
- The fragile foundations of CoT monitoring | Christopher Pottsweb.stanford.edu
- Quantifying the Necessity of Chain of Thought through Opaque Serial Deptharxiv.org
- What is neuralese and why is everyone so concerned about it?transformernews.ai
- Clouded Judgement 9.4.26 - Recurrent Depthcloudedjudgement.substack.com
- Proposal for tracking the effects of architecture on monitorability — Redwood Researchredwoodresearch.org
- GitHub - huskydoge/Awesome-Loop-Models: A curated list of papers and selected technical blogs on Loop Models.github.com
- When Chain of Thought is Necessary, Language Models Struggle to Evade Monitorsarxiv.org