YoutuBlog(HackerTube): YouTube 발표와 팟캐스트를 글로 읽기
필터 결과
46개정렬
16:48EN
Nathan Lambert· 2026년 11월 2일Nathan Lambert on How Open Models Can Compete With Frontier LabsNathan Lambert · Allen Institute for AIOpen models are gaining use and closing capability gaps, but their builders need a sustainable economy and advantages beyond price.
48:04ENNathan Lambert· 2026년 10월 26일Nathan Lambert on Open Language Models: History, Competition and RiskNathan Lambert · Allen Institute for AIA tour of open weights, research infrastructure, distillation, adoption and the choices facing model developers.
46:42ENNathan Lambert· 2026년 10월 25일Nathan Lambert and Florian Brand on Kimi K3.1, Cyber Risk and AI AgentsNathan Lambert, Florian Brand · Allen Institute for AIThe Interconnects hosts discuss mixed model evaluations, a wave of security disclosures, and what a personal agent can do in everyday life.
5:42ENNathan Lambert· 2026년 8월 12일Why the Shoggoth Became a Mascot for AI Post-TrainingNathan Lambert · Allen Institute for AINathan Lambert traces a creature from Lovecraft to a visual metaphor for language models, fine-tuning, and RLHF.
32:56ENNathan Lambert· 2026년 8월 6일Nathan Lambert: An Introduction to AI Character TrainingNathan Lambert · Allen Institute for AIHow post-training shapes a model’s behavior, from constitutions and model specs to experiments with personality and activation steering.
32:11ENNathan Lambert· 2026년 8월 5일Nathan Lambert on How Language Model Evaluation Has EvolvedNathan Lambert · Allen Institute for AIFrom few-shot prompts and pass@k to agent sandboxes, noisy scores and benchmark contamination.
13:53ENNathan Lambert· 2026년 8월 4일Nathan Lambert Answers Six Questions About Post-Training RecipesNathan Lambert · Allen Institute for AIA practical discussion of specialist models, KL penalties, supervised data, reinforcement learning and business tradeoffs.
37:47ENNathan Lambert· 2026년 7월 30일Nathan Lambert on Tool Use and Training Language-Model AgentsNathan Lambert · Allen Institute for AIFrom structured tool calls and harnesses to the data, sandboxes and stability challenges of agentic reinforcement learning.
1:06:30ENNathan Lambert· 2026년 7월 29일Scott Geng and Nathan Lambert: Bringing DPO Research Into OLMo 3Nathan Lambert, Scott Geng · Allen Institute for AIA preference-learning result showed promise in small experiments. Turning it into a model stage required data, organizational and infrastructure work.
31:33ENNathan Lambert· 2026년 7월 28일Nathan Lambert on KL Regularization, RL and Supervised Fine-TuningNathan Lambert · Allen Institute for AIWhy policy updates need control, how KL direction shapes objectives, and what experiments suggest about generalization and forgetting.
23:34ENNathan Lambert· 2026년 7월 25일Nathan Lambert on Over-Optimization and RLHF’s ReputationNathan Lambert · Allen Institute for AIWhy reward-model scores can rise as real behavior worsens, and how preference tuning can still improve useful capabilities.
35:53ENNathan Lambert· 2026년 7월 20일Nathan Lambert on Preference Data and the Foundations of RLHFNathan Lambert · Allen Institute for AIHow ideas from philosophy and control became preference models, and why collecting feedback remains difficult.
14:55ENNathan Lambert· 2026년 7월 1일Nathan Lambert: Fixing RLHF Derivations and Reinforcement-Learning NotationNathan Lambert · Allen Institute for AIA reader-question roundup on KL estimates, reward-model losses, practical experiments and policy gradients.
23:52ENNathan Lambert· 2026년 6월 24일Nathan Lambert: Machine-Learning Foundations for LLM Post-TrainingNathan Lambert · Allen Institute for AIA practical primer on token probabilities, masks, optimization, KL divergence and reinforcement learning.
49:40ENNathan Lambert· 2026년 6월 23일Nathan Lambert on Synthetic Data and On-Policy DistillationNathan Lambert · Allen Institute for AIFrom teacher–student training and KL objectives to Constitutional AI, AI judges and rubric-based rewards.
6:20ENNathan Lambert· 2026년 6월 22일Nathan Lambert on TMax: An Open Training Recipe for Terminal AgentsNathan Lambert · Allen Institute for AIWhy a release of terminal environments, model checkpoints and stable RL infrastructure matters for open agent research.
15:22ENNathan Lambert· 2026년 6월 15일Nathan Lambert Answers Questions on Teacher Models and PPONathan Lambert · Allen Institute for AIWhy the strongest teacher may not transfer best, and how PPO mathematics maps to language-model code.
45:20ENNathan Lambert· 2026년 6월 15일Nathan Lambert on the Rise of Reasoning ModelsNathan Lambert · Allen Institute for AIA history of verifiable-reward training, reasoning-model releases and the implementation choices behind them.
42:44ENNathan Lambert· 2026년 6월 15일Nathan Lambert Explains DPO and Direct Preference OptimizationNathan Lambert · Allen Institute for AIThe derivation, practical limits and related algorithms behind a simpler way to train on preferences.
46:10ENNathan Lambert· 2026년 4월 14일Nathan Lambert: An Overview of Post-Training and RLHFNathan Lambert · Allen Institute for AIFrom language-model pretraining and human preferences to the classic RLHF recipe and scaled reasoning.
3:35ENNathan Lambert· 2026년 4월 14일Nathan Lambert Introduces the RLHF Book Post-Training CourseNathan Lambert · Allen Institute for AIWhy the book and course document practical methods, mathematics and data used to train language models.
53:36ENNathan Lambert· 2026년 4월 14일Nathan Lambert on Implementing RL Algorithms for Language ModelsNathan Lambert · Allen Institute for AIA practical walkthrough of PPO, GRPO, distributed training and the loss-normalization choices that affect model behavior.
57:35ENNathan Lambert· 2026년 4월 14일Nathan Lambert Explains Policy Gradients for Language ModelsNathan Lambert · Allen Institute for AIFrom the policy-gradient theorem and variance reduction to PPO, GRPO and modern variants.
49:49ENNathan Lambert· 2026년 4월 14일Nathan Lambert on Instruction Tuning, Reward Models and Rejection SamplingNathan Lambert · Allen Institute for AIThe foundations of post-training, from chat templates and preference losses to selecting model-generated answers.더 보기46개 중 24개 표시