How language model post-training is done today

array(20) { ["YXTYbr3hiFU"]=> object(stdClass)#7356 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "YXTYbr3hiFU" ["related_video_title"]=> string(48) "An Unexpected Reinforcement Learning Renaissance" ["posted_time"]=> string(27) "9 месяцев назад" ["channelName"]=> NULL } ["FUcilE5Gx_0"]=> object(stdClass)#7365 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "FUcilE5Gx_0" ["related_video_title"]=> string(50) "Состояние открытых моделей" ["posted_time"]=> string(25) "4 недели назад" ["channelName"]=> NULL } ["ZkYNjV1qETk"]=> object(stdClass)#7354 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "ZkYNjV1qETk" ["related_video_title"]=> string(131) "Если это сработает… ИИ появится раньше. (Термодинамические вычисления)" ["posted_time"]=> string(23) "8 часов назад" ["channelName"]=> NULL } ["wjZofJX0v4M"]=> object(stdClass)#7359 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "wjZofJX0v4M" ["related_video_title"]=> string(148) "LLM и GPT - как работают большие языковые модели? Визуальное введение в трансформеры" ["posted_time"]=> string(19) "1 год назад" ["channelName"]=> NULL } ["yGkJj_4bjpE"]=> object(stdClass)#7348 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "yGkJj_4bjpE" ["related_video_title"]=> string(77) "How I finetuned a Small LM to THINK and solve puzzles on its own (GRPO & RL!)" ["posted_time"]=> string(25) "4 месяца назад" ["channelName"]=> NULL } ["VAzL8RHot1c"]=> object(stdClass)#7364 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "VAzL8RHot1c" ["related_video_title"]=> string(53) "The art of training a good (reasoning) language model" ["posted_time"]=> string(25) "4 месяца назад" ["channelName"]=> NULL } ["H-oCV5brtU4"]=> object(stdClass)#7352 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "H-oCV5brtU4" ["related_video_title"]=> string(42) "Intro to Fine-Tuning Large Language Models" ["posted_time"]=> string(25) "2 месяца назад" ["channelName"]=> NULL } ["1pmyTnGOevU"]=> object(stdClass)#7360 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "1pmyTnGOevU" ["related_video_title"]=> string(56) "Introduction to LLM Post Training by Maxime Labonne, PhD" ["posted_time"]=> string(23) "1 месяц назад" ["channelName"]=> NULL } ["DvZ8jZ-laj4"]=> object(stdClass)#7342 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "DvZ8jZ-laj4" ["related_video_title"]=> string(89) "Reasoning without Language - Deep Dive into 27 mil parameter Hierarchical Reasoning Model" ["posted_time"]=> string(25) "3 месяца назад" ["channelName"]=> NULL } ["vIgE1t1rKjg"]=> object(stdClass)#7363 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "vIgE1t1rKjg" ["related_video_title"]=> string(33) "They Built an AGI Lab in 8 Months" ["posted_time"]=> string(19) "3 дня назад" ["channelName"]=> NULL } ["C4HxJQ2QzWo"]=> object(stdClass)#7355 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "C4HxJQ2QzWo" ["related_video_title"]=> string(39) "Reinforcement Learning for LLMs in 2025" ["posted_time"]=> string(27) "9 месяцев назад" ["channelName"]=> NULL } ["pkpJMNjvgXw"]=> object(stdClass)#7361 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "pkpJMNjvgXw" ["related_video_title"]=> string(57) "David Silver - Towards Superhuman Intelligence - RLC 2024" ["posted_time"]=> string(19) "1 год назад" ["channelName"]=> NULL } ["J1APR8Bo9dE"]=> object(stdClass)#7349 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "J1APR8Bo9dE" ["related_video_title"]=> string(65) "Early stages of the reinforcement learning era of language models" ["posted_time"]=> string(27) "8 месяцев назад" ["channelName"]=> NULL } ["KJtZARuO3JY"]=> object(stdClass)#7347 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "KJtZARuO3JY" ["related_video_title"]=> string(70) "Visualizing transformers and attention | Talk for TNG Big Tech Day '24" ["posted_time"]=> string(28) "11 месяцев назад" ["channelName"]=> NULL } ["LVXtFnEbNU0"]=> object(stdClass)#7345 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "LVXtFnEbNU0" ["related_video_title"]=> string(100) "Everything You Wanted to Know About LLM Post-Training, with Nathan Lambert of Allen Institute for AI" ["posted_time"]=> string(28) "11 месяцев назад" ["channelName"]=> NULL } ["oogxOPjGL04"]=> object(stdClass)#7346 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "oogxOPjGL04" ["related_video_title"]=> string(167) "Россия просто исчезает. Предупреждение, к которому нужно прислушаться / Валентин Катасонов" ["posted_time"]=> string(23) "9 часов назад" ["channelName"]=> NULL } ["-8zrQggsVJU"]=> object(stdClass)#7343 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "-8zrQggsVJU" ["related_video_title"]=> string(180) "«Я понял, что это конец»: как создатель «Алисы» уволился из «Сбера», эмигрировал и строит AI-стартап" ["posted_time"]=> string(25) "2 недели назад" ["channelName"]=> NULL } ["eTieetk2dSw"]=> object(stdClass)#7344 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "eTieetk2dSw" ["related_video_title"]=> string(58) "Building with Instruction-Tuned LLMs: A Step-by-Step Guide" ["posted_time"]=> string(65) "Трансляция закончилась 2 года назад" ["channelName"]=> NULL } ["g80Q1sVtikE"]=> object(stdClass)#7332 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "g80Q1sVtikE" ["related_video_title"]=> string(108) "Доработайте свою степень магистра права за 13 минут. Вот как" ["posted_time"]=> string(23) "1 месяц назад" ["channelName"]=> NULL } ["PAz_-xPJcRM"]=> object(stdClass)#7333 (5) { ["video_id"]=> int(9999999) ["related_video_id"]=> string(11) "PAz_-xPJcRM" ["related_video_title"]=> string(0) "" ["posted_time"]=> string(25) "3 месяца назад" ["channelName"]=> NULL } }