Lawrence Feng

on the atlas — 22

• It's nice of you to worry about me, but I really do have a life — LessWrong1 savers
• Unfamiliar Finetuning Examples Control How Language Models Hallucinate1 savers
• Anthropic's leading researchers acted as moderate accelerationists — LessWrong1 savers
• Reinforcement Learning Finetunes Small Subnetworks in Large Language Models1 savers
• LoRA vs Full Fine-tuning: An Illusion of Equivalence1 savers
• Lawrence's Bookshelf / Curius1 savers
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning1 savers
• When Chain of Thought is Necessary, Language Models Struggle to Evade Monitors1 savers
• Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation1 savers
• LoRA Without Regret - Thinking Machines Lab32 savers
• The machines are fine. I'm worried about us.22 savers
• Did Claude 3 Opus align itself via gradient hacking? — LessWrong16 savers
• Alignment remains a hard, unsolved problem — LessWrong14 savers
• Responsible Scaling Policy v3 — LessWrong7 savers
• Intentionally Designing the Future of AI7 savers
• Where I agree and disagree with Eliezer — LessWrong6 savers
• How Go Players Disempower Themselves to AI — LessWrong6 savers
• Some Math behind Neural Tangent Kernel | Lil'Log5 savers
• Home - Joe Carlsmith4 savers
• The shard theory of human values - LessWrong4 savers
• Believe It or Not: How Deeply do LLMs Believe Implanted Facts?3 savers
• Andrej Karpathy on X: "LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating" / X3 savers