正在加载视频...

视频加载失败

返回首页

Reasoning from scratch, round number 5! This time, talking about log-probability scoring (also a great fundamental concept for loss functions like cross-entropy in pre-training and distillation) and self-refinement. 00:00 Introduction and inference-time scaling recap 05:02 Loading the pretrained LLM 08:00 Comparing and scoring model answers 10:18 Building a rule-based...

rasbt's profile picture

Sebastian Raschka

529,010 subscribers

59,636 次观看 • 1 个月前 •via X (Twitter)

科学技术
LIVE

Anya Rossi• Live Now

Private livecam show

27 条评论

Sebastian Raschka 的头像
Sebastian Raschka1 个月前

And a link to the video on YouTube:

Parul Pandey 的头像
Parul Pandey1 个月前

Reading this chapter from your book as I type. Its a joy to read it. Also the visuals make it really easy to grasp the topic.

Sebastian Raschka 的头像
Sebastian Raschka1 个月前

Thanks a lot! Put lots of work went into it, and it really feels good to hear this :)

Giacomo Masseroni 的头像
Giacomo Masseroni1 个月前

log probs for numerical stability instead of raw probabilities, that part alone saves people from nan hell

Sebastian Raschka 的头像
Sebastian Raschka1 个月前

Yup. Also covered that explicitly there :)

Victor Akinode 的头像
Victor Akinode1 个月前

The timestamps make this much easier to learn from, Sebastian. For anyone following along, changing one scoring choice and comparing the answers could be a lovely little experiment. It gives those probability calculations something concrete to connect to.

Sebastian Raschka 的头像
Sebastian Raschka1 个月前

Thanks!

Fireply.ai 的头像
Fireply.ai1 个月前

timestamps going past an hour and it's still just round 5, this book is going to be dense

Sebastian Raschka 的头像
Sebastian Raschka1 个月前

some would say "you get your money's worth" :P

lewington 的头像
lewington1 个月前

log probabilities are quite useful to understand cuz getting a model to generate an answer and figuring out if its answer is actually better are two very different things.

Daniel Priscu 的头像
Daniel Priscu1 个月前

log-prob scoring pairs well with self-refinement, since the model can rank its own revisions instead of guessing which one improved

The Lunar Misfit 的头像
The Lunar Misfit1 个月前

Self-refinement is a much kinder name than 'read that back to yourself'.

David Arnal 的头像
David Arnal1 个月前

The average log-probability scoring setup is a nice detail; it also makes the numerical-stability discussion feel directly relevant rather than incidental.

Siddharth Jain 的头像
Siddharth Jain1 个月前

Computing token probabilities in PyTorch was my fav part, very well explained.

Bk1man｜AI 开源雷达 的头像
Bk1man｜AI 开源雷达1 个月前

想快速验证自精炼到底贡献了多少，可以先做一个 20 分钟测试：① 准备 50 道有标准答案的题先跑一遍，记下基线准确率 ② 打开自精炼再跑一遍，统计修正率和误修率 ③ 把自精炼关掉、只保留 logprob 阈值过滤跑第三遍，看阈值这一步单独贡献了多少。这样就知道收益是来自精炼本身，还是来自筛掉低置信样本。

Jatin Garg 的头像
Jatin Garg1 个月前

Log-probability scoring makes sense for evaluating model confidence, but how do you handle cases where the model is confidently wrong? High log-prob on an incorrect answer seems like it would still score well.

Sebastian Raschka 的头像
Sebastian Raschka1 个月前

Yes, this is confidence, not correctness.I am talking through that issue in this video. You could technically use a second stronger model though that checks the answer. Kind of like speculative decoding works (the more expensive model accepts/rejects cheaply generated suggestions). But yeah, either way, the other motivation to introduce log-probability scoring is that it is essential for the upcoming training chapters (RL and distillations).

VisiveAI 的头像
VisiveAI1 个月前

Log-prob scoring is the boring fundamental that keeps evals honest — glad R5 drills it from scratch. Stack ↓ #LLM #ML

Saksham Jain 的头像
Saksham Jain1 个月前

does log-prob self-scoring hold up when the model's miscalibrated? wondered if self-refinement just reinforces the confident wrong ones

Oleks 的头像
Oleks1 个月前

saving the log-prob section for tomorrow morning

dhinna ship .ico 的头像
dhinna ship .ico1 个月前

Token-level or sequence-level log-probs in the self-refinement loop?

abdul samad 的头像
abdul samad1 个月前

You have been source of learning for millions, thank you very much for creating these videos

Sebastian Raschka 的头像
Sebastian Raschka1 个月前

thank you for the kind words!

VastPlan 的头像
VastPlan1 个月前

对数概率评分把候选答案压成可比较的序列似然，也是交叉熵与蒸馏损失的共用底座。把规则打分换成平均对数概率后，自我精炼循环才有稳定的排序依据；若评论器与生成器共用同一套偏差，MATH-500 上的增益仍需与独立裁判对照。

Elliott 的头像
Elliott1 个月前

I am going to watch the heck out of this Seb

dish 的头像
dish1 个月前

Great point that log-probs measure confidence, not correctness. High confidence on a wrong answer is the dangerous case. How do you calibrate or downweight that in a self-refinement loop?

ethereagle · building 的头像
ethereagle · building1 个月前

self-refinement only pays if the scorer already agrees with MATH-500. did average logprob ranking move the score on its own, or only after the critique loop?

相关视频

Reasoning from scratch, round number 6! An introduction (and implementation) of Reinforcement Learning with Verifiable Rewards (RLVR) and Group Relative Policy Optimization (GRPO). 00:00 Introduction 01:54 What makes a reasoning model different? 04:25 Reasoning traces and model capability 08:29 Accuracy and format rewards 11:34 Aha moments and DeepSeek-R1 training 14:41 Reasoning effort and answer length 18:38 RLHF and RLVR 23:04 GRPO vs. PPO 26:40 GRPO explained with a cooking analogy 31:43 The KL term and simplified GRPO 35:04 Loading the pretrained model 36:07 Loading the MATH training data 39:26 Sampling model responses 46:30 Computing verifiable rewards 49:55 Computing advantages 51:54 Token and sequence log probabilities 55:29 Implementing sequence log probabilities 57:37 Fixing the inference-mode error 1:02:24 Computing the GRPO loss 1:04:37 Putting the GRPO step together 1:09:19 The GRPO training loop 1:12:57 Training settings, logging, and checkpoints 1:17:24 Running training and inspecting outputs 1:19:28 Loading and evaluating checkpoints 1:22:33 MATH-500 results and training stability 1:24:05 Memory requirements and next steps
1:26:53

Reasoning from scratch, round number 6! An introduction (and implementation) of Reinforcement Learning with Verifiable Rewards (RLVR) and Group Relative Policy Optimization (GRPO). 00:00 Introduction 01:54 What makes a reasoning model different? 04:25 Reasoning traces and model capability 08:29 Accuracy and format rewards 11:34 Aha moments and DeepSeek-R1 training 14:41 Reasoning effort and answer length 18:38 RLHF and RLVR 23:04 GRPO vs. PPO 26:40 GRPO explained with a cooking analogy 31:43 The KL term and simplified GRPO 35:04 Loading the pretrained model 36:07 Loading the MATH training data 39:26 Sampling model responses 46:30 Computing verifiable rewards 49:55 Computing advantages 51:54 Token and sequence log probabilities 55:29 Implementing sequence log probabilities 57:37 Fixing the inference-mode error 1:02:24 Computing the GRPO loss 1:04:37 Putting the GRPO step together 1:09:19 The GRPO training loop 1:12:57 Training settings, logging, and checkpoints 1:17:24 Running training and inspecting outputs 1:19:28 Loading and evaluating checkpoints 1:22:33 MATH-500 results and training stability 1:24:05 Memory requirements and next steps

Sebastian Raschka

76,474 次观看 • 1 个月前

Reasoning From Scratch round 8, the finale! This last lecture in the series covers distillation as an efficient alternative to the previous reinforcement learning training runs. (Planning to take a brief break, but I will be back!) 00:00 Introduction and series recap 01:07 Why distill a reasoning model? 02:51 Soft vs. hard distillation 11:06 On-policy vs. off-policy distillation 18:09 Preparing data for reasoning distillation 22:35 Loading the DeepSeek R1 MATH dataset 28:04 Inspecting reasoning traces and final answers 31:27 Tokenization and chat templates 35:58 Pre-tokenizing the training dataset 39:01 Sequence lengths and the 2,048-token limit 42:38 Creating training and validation splits 45:32 Loading the Qwen3-0.6B base model 46:24 Computing answer-token log probabilities 54:10 Cross-entropy loss for distillation 57:25 Training and validation loss helpers 1:02:03 Implementing the distillation training loop 1:07:26 Running the standalone training script 1:09:41 Plotting training and validation losses 1:11:31 Evaluating checkpoints on MATH-500 1:14:09 Comparing teachers and distillation results 1:16:34 Future directions for reasoning models 1:19:41 Bonus exercises and next steps
1:21:17

Reasoning From Scratch round 8, the finale! This last lecture in the series covers distillation as an efficient alternative to the previous reinforcement learning training runs. (Planning to take a brief break, but I will be back!) 00:00 Introduction and series recap 01:07 Why distill a reasoning model? 02:51 Soft vs. hard distillation 11:06 On-policy vs. off-policy distillation 18:09 Preparing data for reasoning distillation 22:35 Loading the DeepSeek R1 MATH dataset 28:04 Inspecting reasoning traces and final answers 31:27 Tokenization and chat templates 35:58 Pre-tokenizing the training dataset 39:01 Sequence lengths and the 2,048-token limit 42:38 Creating training and validation splits 45:32 Loading the Qwen3-0.6B base model 46:24 Computing answer-token log probabilities 54:10 Cross-entropy loss for distillation 57:25 Training and validation loss helpers 1:02:03 Implementing the distillation training loop 1:07:26 Running the standalone training script 1:09:41 Plotting training and validation losses 1:11:31 Evaluating checkpoints on MATH-500 1:14:09 Comparing teachers and distillation results 1:16:34 Future directions for reasoning models 1:19:41 Bonus exercises and next steps

Sebastian Raschka

39,937 次观看 • 27 天前

Inference scaling part 1. Starting with a modded text generation function (temperature scaling, top-p filtering, multinomial sampling) to generate diverse outputs for self-consistency and best-of-N (improving answer accuracy by>2x) 00:00 Introduction and recap 00:31 Training-time and inference-time scaling 07:52 What we'll implement 11:47 Notebook setup and model loading 17:43 Building a flexible text generation function 24:40 Chain-of-thought prompting 28:26 Sampling and output diversity 33:43 Next-token logits and greedy decoding 38:20 Temperature scaling step by step 42:46 Softmax and token probabilities 47:42 Multinomial sampling 54:51 Adding temperature sampling to text generation 59:31 Top-p filtering step by step 1:10:23 Adding top-p filtering to text generation 1:13:43 Sampling and LLM watermarking 1:16:01 Self-consistency and majority voting 1:20:36 Implementing self-consistency 1:29:02 MATH-500 results 1:35:01 Accuracy and compute tradeoffs 1:36:50 Next steps and self-refinement
1:37:11

Inference scaling part 1. Starting with a modded text generation function (temperature scaling, top-p filtering, multinomial sampling) to generate diverse outputs for self-consistency and best-of-N (improving answer accuracy by>2x) 00:00 Introduction and recap 00:31 Training-time and inference-time scaling 07:52 What we'll implement 11:47 Notebook setup and model loading 17:43 Building a flexible text generation function 24:40 Chain-of-thought prompting 28:26 Sampling and output diversity 33:43 Next-token logits and greedy decoding 38:20 Temperature scaling step by step 42:46 Softmax and token probabilities 47:42 Multinomial sampling 54:51 Adding temperature sampling to text generation 59:31 Top-p filtering step by step 1:10:23 Adding top-p filtering to text generation 1:13:43 Sampling and LLM watermarking 1:16:01 Self-consistency and majority voting 1:20:36 Implementing self-consistency 1:29:02 MATH-500 results 1:35:01 Accuracy and compute tradeoffs 1:36:50 Next steps and self-refinement

Sebastian Raschka

44,580 次观看 • 1 个月前

Reasoning from scratch round 3: This time, I cover generating a verifier for... a) ...evaluation (base model versus any future model improvement) b) ...the reinforcement learning with verifiable rewards (RLVR) training later on 00:00 Introduction 01:21 Four approaches to LLM evaluation 07:20 Verifiers and reinforcement learning with verifiable rewards 10:52 Notebook setup and dependencies 13:43 Section 3.1 Building a math verifier 18:57 Section 3.2 Loading a pre-trained model to generate text 24:34 Generating and displaying model answers 29:23 Section 3.3 Implementing a wrapper for easier text generation 34:00 Section 3.4 Extracting the final answer box 37:29 Handling answers without boxes 43:17 Section 3.5 Normalizing the extracted answer 46:56 Section 3.6 Verifying mathematical equivalence 53:32 Implementing the equality check 57:48 Section 3.7 Grading answers 59:20 Building and testing the answer grader 1:03:18 Section 3.8 Loading the evaluation dataset (MATH-500) 1:07:51 Section 3.9 Evaluating the model 1:08:34 Prompt templates for evaluation 1:10:47 Prompt sensitivity and memorization 1:13:55 A minimal evaluation example 1:15:32 Building the evaluation loop 1:20:27 Comparing CPU, MPS, and CUDA results 1:21:54 Reproducibility and floating-point math 1:23:37 Base model vs. reasoning model 1:25:30 Summary and next steps
1:26:46

Reasoning from scratch round 3: This time, I cover generating a verifier for... a) ...evaluation (base model versus any future model improvement) b) ...the reinforcement learning with verifiable rewards (RLVR) training later on 00:00 Introduction 01:21 Four approaches to LLM evaluation 07:20 Verifiers and reinforcement learning with verifiable rewards 10:52 Notebook setup and dependencies 13:43 Section 3.1 Building a math verifier 18:57 Section 3.2 Loading a pre-trained model to generate text 24:34 Generating and displaying model answers 29:23 Section 3.3 Implementing a wrapper for easier text generation 34:00 Section 3.4 Extracting the final answer box 37:29 Handling answers without boxes 43:17 Section 3.5 Normalizing the extracted answer 46:56 Section 3.6 Verifying mathematical equivalence 53:32 Implementing the equality check 57:48 Section 3.7 Grading answers 59:20 Building and testing the answer grader 1:03:18 Section 3.8 Loading the evaluation dataset (MATH-500) 1:07:51 Section 3.9 Evaluating the model 1:08:34 Prompt templates for evaluation 1:10:47 Prompt sensitivity and memorization 1:13:55 A minimal evaluation example 1:15:32 Building the evaluation loop 1:20:27 Comparing CPU, MPS, and CUDA results 1:21:54 Reproducibility and floating-point math 1:23:37 Base model vs. reasoning model 1:25:30 Summary and next steps

Sebastian Raschka

43,229 次观看 • 2 个月前

Reasoning from scratch round 2: In this video, I cover the text generation process in LLMs and KV caching (to prepare the base model before adding reasoning techniques in the upcoming ones). 00:00 Introduction and reasoning model demo 01:55 How to work through the book 05:00 Chapter 2 overview 08:25 Checking PyTorch and hardware support 10:26 Apple silicon and MPS caveats 15:00 Cloud GPU options 16:08 Tokens and tokenization 18:20 Qwen3 and the Reasoning From Scratch package 23:05 Encoding and decoding text 26:24 Downloading weights and selecting a device 31:01 Loading the pretrained Qwen3 model 34:32 How LLMs generate text 36:47 Input tensors and batch dimensions 41:48 Running the model in inference mode 44:11 Logits and next-token predictions 49:21 Greedy decoding with argmax 52:28 Building a streaming text generator 01:01:28 Generating text and handling end-of-sequence tokens 01:06:00 Benchmarking text generation 01:14:34 How KV caching works 01:17:22 Adding KV caching and measuring the speedup 01:24:31 Model compilation with torch.compile 01:30:33 Combining compilation with KV caching 01:32:53 Comparing CPU and GPU performance 01:35:32 Recap and next steps
1:36:42

Reasoning from scratch round 2: In this video, I cover the text generation process in LLMs and KV caching (to prepare the base model before adding reasoning techniques in the upcoming ones). 00:00 Introduction and reasoning model demo 01:55 How to work through the book 05:00 Chapter 2 overview 08:25 Checking PyTorch and hardware support 10:26 Apple silicon and MPS caveats 15:00 Cloud GPU options 16:08 Tokens and tokenization 18:20 Qwen3 and the Reasoning From Scratch package 23:05 Encoding and decoding text 26:24 Downloading weights and selecting a device 31:01 Loading the pretrained Qwen3 model 34:32 How LLMs generate text 36:47 Input tensors and batch dimensions 41:48 Running the model in inference mode 44:11 Logits and next-token predictions 49:21 Greedy decoding with argmax 52:28 Building a streaming text generator 01:01:28 Generating text and handling end-of-sequence tokens 01:06:00 Benchmarking text generation 01:14:34 How KV caching works 01:17:22 Adding KV caching and measuring the speedup 01:24:31 Model compilation with torch.compile 01:30:33 Combining compilation with KV caching 01:32:53 Comparing CPU and GPU performance 01:35:32 Recap and next steps

Sebastian Raschka

68,300 次观看 • 2 个月前

Reasoning From Scratch: Reinforcement Learning with Verifiable Rewards (RLVR) round 2. Covering clipped policy ratios, KL loss term, format rewards, and other GRPO tips & tricks. 00:00 Introduction and recap 01:52 Interpreting basic GRPO training metrics 06:34 Planned improvements to GRPO 08:58 Running longer training jobs with Python scripts 13:39 Running the baseline GRPO training script 17:29 Loading and plotting training logs 19:29 Diagnosing unstable training 23:55 Evaluating checkpoints on MATH-500 26:26 Downloading existing checkpoints 30:09 Tracking advantage statistics 34:53 Understanding entropy 40:32 Computing entropy in PyTorch 44:17 Interpreting entropy values 48:58 Adding entropy tracking to GRPO 53:36 Analyzing advantage and entropy metrics 56:18 Stabilizing GRPO with clipped policy ratios 1:03:27 Implementing the clipped policy loss 1:09:39 Analyzing clipped policy training results 1:11:25 KL divergence and reward hacking 1:15:12 Adding a KL loss term 1:20:34 Limitations of the simplified KL loss 1:23:04 Format rewards and think tags 1:25:47 Adding special tokens to the tokenizer 1:30:29 Implementing the format reward 1:35:56 Analyzing format reward training 1:38:25 Rewarding format only for correct answers 1:40:48 Further GRPO improvements from research 1:45:43 Next steps and distillation
1:46:07

Reasoning From Scratch: Reinforcement Learning with Verifiable Rewards (RLVR) round 2. Covering clipped policy ratios, KL loss term, format rewards, and other GRPO tips & tricks. 00:00 Introduction and recap 01:52 Interpreting basic GRPO training metrics 06:34 Planned improvements to GRPO 08:58 Running longer training jobs with Python scripts 13:39 Running the baseline GRPO training script 17:29 Loading and plotting training logs 19:29 Diagnosing unstable training 23:55 Evaluating checkpoints on MATH-500 26:26 Downloading existing checkpoints 30:09 Tracking advantage statistics 34:53 Understanding entropy 40:32 Computing entropy in PyTorch 44:17 Interpreting entropy values 48:58 Adding entropy tracking to GRPO 53:36 Analyzing advantage and entropy metrics 56:18 Stabilizing GRPO with clipped policy ratios 1:03:27 Implementing the clipped policy loss 1:09:39 Analyzing clipped policy training results 1:11:25 KL divergence and reward hacking 1:15:12 Adding a KL loss term 1:20:34 Limitations of the simplified KL loss 1:23:04 Format rewards and think tags 1:25:47 Adding special tokens to the tokenizer 1:30:29 Implementing the format reward 1:35:56 Analyzing format reward training 1:38:25 Rewarding format only for correct answers 1:40:48 Further GRPO improvements from research 1:45:43 Next steps and distillation

Sebastian Raschka

60,891 次观看 • 1 个月前

François Chollet (François Chollet) on the ARC Prize and how we get to AGI. At AI Startup School in San Francisco. 00:00 - The Falling Cost of Compute 00:57 - Deep-Learning’s Scaling Era & Benchmarks 01:59 - The ARC Benchmark 03:02 - The 2024 Shift to Test-Time Adaptation 05:01 - What Is Intelligence? 07:12 - Why Benchmarks Matter (and Mislead) 08:57 - ARC 1 Exposes Scaling Limits 10:58 - ARC 2: Compositional Reasoning Arrives 12:55 - Humans vs. Models on ARC2 14:58 - Previewing ARC3 & Interactive Agency 17:00 - Kaleidoscopic Hypothesis and Abstractions 22:00 - Type 1 vs. Type 2 Abstractions 26:00 - Discrete Program Search & Inventive AI 29:00 - Fusing Intuition with Symbolic Reasoning 32:00 - Building AGI Through Meta-Learning Systems
34:47

François Chollet (François Chollet) on the ARC Prize and how we get to AGI. At AI Startup School in San Francisco. 00:00 - The Falling Cost of Compute 00:57 - Deep-Learning’s Scaling Era & Benchmarks 01:59 - The ARC Benchmark 03:02 - The 2024 Shift to Test-Time Adaptation 05:01 - What Is Intelligence? 07:12 - Why Benchmarks Matter (and Mislead) 08:57 - ARC 1 Exposes Scaling Limits 10:58 - ARC 2: Compositional Reasoning Arrives 12:55 - Humans vs. Models on ARC2 14:58 - Previewing ARC3 & Interactive Agency 17:00 - Kaleidoscopic Hypothesis and Abstractions 22:00 - Type 1 vs. Type 2 Abstractions 26:00 - Discrete Program Search & Inventive AI 29:00 - Fusing Intuition with Symbolic Reasoning 32:00 - Building AGI Through Meta-Learning Systems

Y Combinator

232,129 次观看 • 1 年前

Reiner Pope (MatX) just raised a $500m round led by Leopold Aschenbrenner and Jane Street to build faster AI chips. I enjoyed having him on Cheeky Pint so I could ask all my questions about how chip design actually works, where the speed-up comes from, and how the industry will evolve. 00:00:15 Google’s AI revival 00:07:54 MatX 00:17:11 AI supply chain 00:21:48 Designing chips 00:37:11 TSMC 00:44:17 Token pricing 00:44:55 RL-ing chip design 00:49:26 Design to production 00:56:05 MatX culture 01:02:57 Rust 01:05:21 Cuckoo hashing 01:09:35 Unexplored model architectures
1:13:04

Reiner Pope (MatX) just raised a $500m round led by Leopold Aschenbrenner and Jane Street to build faster AI chips. I enjoyed having him on Cheeky Pint so I could ask all my questions about how chip design actually works, where the speed-up comes from, and how the industry will evolve. 00:00:15 Google’s AI revival 00:07:54 MatX 00:17:11 AI supply chain 00:21:48 Designing chips 00:37:11 TSMC 00:44:17 Token pricing 00:44:55 RL-ing chip design 00:49:26 Design to production 00:56:05 MatX culture 01:02:57 Rust 01:05:21 Cuckoo hashing 01:09:35 Unexplored model architectures

John Collison

54,548 次观看 • 8 个月前

00:00:00 Introduction and guest overview 00:02:42 Lucas Shuermann and Variational background 00:05:14 Growth and benefits of Omni 00:05:45 How Omni offers tighter spreads and zero fees 00:09:35 Flow segmentation and liquidity quality 00:14:05 Hedging challenges on other venues 00:17:06 Evolution of trading venues and infrastructure 00:18:27 APIs and handling toxic flow 00:21:00 Growth and inefficiencies in real world asset perps (RWA) 00:23:37 Comparing traditional and on-chain liquidity 00:25:29 Challenges of building liquidity from scratch 00:27:31 Aggregation vs rebuilding liquidity 00:28:35 Hedging strategies across crypto and TradFi 00:31:16 Phases of RWA rollout on Omni 00:36:39 Introduction to swaps and their advantages 00:37:57 Setting funding rates for perps 00:41:03 Index change modes and design decisions 00:44:49 Introduction and comparison of swaps 00:48:11 Swaps vs perps, CFDs, and TRSs 00:50:43 User decisions between perps and swaps 00:53:39 Financing rates for swaps 00:58:40 TradFi dealer capacity and asset classes 01:01:53 On-chain vs crypto asset exposure 01:02:45 Constraints to long-term vision for RWAs on-chain 01:04:11 Global market expansion and connectivity 01:06:25 Institutional platform ambitions 01:08:11 Institutional benefits of on-chain swaps 01:10:28 Outside of work obsessions 01:13:05 Closing remarks
1:13:17

00:00:00 Introduction and guest overview 00:02:42 Lucas Shuermann and Variational background 00:05:14 Growth and benefits of Omni 00:05:45 How Omni offers tighter spreads and zero fees 00:09:35 Flow segmentation and liquidity quality 00:14:05 Hedging challenges on other venues 00:17:06 Evolution of trading venues and infrastructure 00:18:27 APIs and handling toxic flow 00:21:00 Growth and inefficiencies in real world asset perps (RWA) 00:23:37 Comparing traditional and on-chain liquidity 00:25:29 Challenges of building liquidity from scratch 00:27:31 Aggregation vs rebuilding liquidity 00:28:35 Hedging strategies across crypto and TradFi 00:31:16 Phases of RWA rollout on Omni 00:36:39 Introduction to swaps and their advantages 00:37:57 Setting funding rates for perps 00:41:03 Index change modes and design decisions 00:44:49 Introduction and comparison of swaps 00:48:11 Swaps vs perps, CFDs, and TRSs 00:50:43 User decisions between perps and swaps 00:53:39 Financing rates for swaps 00:58:40 TradFi dealer capacity and asset classes 01:01:53 On-chain vs crypto asset exposure 01:02:45 Constraints to long-term vision for RWAs on-chain 01:04:11 Global market expansion and connectivity 01:06:25 Institutional platform ambitions 01:08:11 Institutional benefits of on-chain swaps 01:10:28 Outside of work obsessions 01:13:05 Closing remarks

Corey Hoffstein 🏴‍☠️

20,210 次观看 • 2 个月前

유우시 춤 00:00 00:13 00:28 00:57 01:37 01:53 02:13 02:35 02:45 03:01 03:17 03:26 03:41 04:07
4:30

유우시 춤 00:00 00:13 00:28 00:57 01:37 01:53 02:13 02:35 02:45 03:01 03:17 03:26 03:41 04:07

우닷

57,007 次观看 • 1 年前

the next MetaDAO ICO is Solomon Labs. a project loved by VCs, but choosing the futarchy route watch the video to find out what they are building, the team, the competition, and all details about the ICO Timestamps 00:00 Introduction to Solomon Labs 00:28 Understanding the Two-Token Model 00:53 Yield Generation Mechanism 01:33 Market Fit and Target Audience 02:35 Competitor Analysis 06:05 Team Behind Solomon Labs 08:46 ICO Details and Token Distribution 11:49 Final Thoughts and Next Steps
14:00

the next MetaDAO ICO is Solomon Labs. a project loved by VCs, but choosing the futarchy route watch the video to find out what they are building, the team, the competition, and all details about the ICO Timestamps 00:00 Introduction to Solomon Labs 00:28 Understanding the Two-Token Model 00:53 Yield Generation Mechanism 01:33 Market Fit and Target Audience 02:35 Competitor Analysis 06:05 Team Behind Solomon Labs 08:46 ICO Details and Token Distribution 11:49 Final Thoughts and Next Steps

Seb Monty

48,390 次观看 • 1 年前

When AI builds AI: my conversation with Richard Socher about RSI and scientific progress. 00:00 Intro: AI That Improves Itself 00:55 Why Scientific Progress Is Slowing 03:08 The Labyrinth of Human Knowledge 05:59 Can AI Put Science Back Together? 07:57 How LLMs Learn Biology and Proteins 10:56 Next-Token Prediction as a World Model 16:44 Can AI Generate Truly Original Ideas? 17:32 Simulations, Verifiers and Superhuman AI 22:18 The Path to Recursive Self-Improvement 24:49 Why AI Hallucinations Can Drive Discovery 27:42 From Reading Biology to Writing It 31:31 Can AI Accelerate Drug Discovery? 33:31 Will AI Help Cure Cancer? 38:03 AI Breakthroughs in Biology, Energy and Materials 40:19 Will Some Societies Reject AI? 45:07 Building the AI Economist 52:22 The Scientific Data Bottleneck 53:41 The Four Pillars of the Eureka Machine 55:01 Teaching AI the Rules of Reality 57:44 Simulations and Virtual Cells 1:00:40 Self-Driving Robotic Laboratories 1:02:51 Agent Swarms and Open-Ended Discovery 1:04:30 The Compute Bottleneck 1:05:44 Inside Recursive 1:07:33 What Recursive Will Build First 1:10:10 How Do We Define Intelligence? 1:11:32 How Far Can Intelligence Go?
1:14:15

When AI builds AI: my conversation with Richard Socher about RSI and scientific progress. 00:00 Intro: AI That Improves Itself 00:55 Why Scientific Progress Is Slowing 03:08 The Labyrinth of Human Knowledge 05:59 Can AI Put Science Back Together? 07:57 How LLMs Learn Biology and Proteins 10:56 Next-Token Prediction as a World Model 16:44 Can AI Generate Truly Original Ideas? 17:32 Simulations, Verifiers and Superhuman AI 22:18 The Path to Recursive Self-Improvement 24:49 Why AI Hallucinations Can Drive Discovery 27:42 From Reading Biology to Writing It 31:31 Can AI Accelerate Drug Discovery? 33:31 Will AI Help Cure Cancer? 38:03 AI Breakthroughs in Biology, Energy and Materials 40:19 Will Some Societies Reject AI? 45:07 Building the AI Economist 52:22 The Scientific Data Bottleneck 53:41 The Four Pillars of the Eureka Machine 55:01 Teaching AI the Rules of Reality 57:44 Simulations and Virtual Cells 1:00:40 Self-Driving Robotic Laboratories 1:02:51 Agent Swarms and Open-Ended Discovery 1:04:30 The Compute Bottleneck 1:05:44 Inside Recursive 1:07:33 What Recursive Will Build First 1:10:10 How Do We Define Intelligence? 1:11:32 How Far Can Intelligence Go?

Matt Turck

10,849 次观看 • 2 个月前

Another quick lecture -- I've been asked many times for prereq's to my book and what you should know, so built a little lecture (with GLM 5.2) to cover some more basics. Topics include: 00:00 Introduction & Course Prerequisites 01:37 Language Models Overview 02:47 The LM Head 04:29 Softmax & Log-Probabilities 06:13 Anatomy of an LM Training Example 06:37 Computing LLM Probabilities (+Phoebe the Dog) 09:52 Three Common Masks in Post-Training 11:03 A Small Decoding Review 12:14 Training an LM: Cross-Entropy 13:23 Optimization & Fine-Tuning 13:55 Pretraining to Midtraining to SFT Pipeline 15:25 Probability Essentials: KL Divergence & Entropy 19:36 Sigmoid & Pairwise Likelihood 20:29 Reinforcement Learning Framing (MDP) 22:28 Transitioning Tools into Post-Training 23:12 Recommended Resources & Wrap-Up  Happy learning and I'm still taking questions from during the course for Q&A videos.
23:52

Another quick lecture -- I've been asked many times for prereq's to my book and what you should know, so built a little lecture (with GLM 5.2) to cover some more basics. Topics include: 00:00 Introduction & Course Prerequisites 01:37 Language Models Overview 02:47 The LM Head 04:29 Softmax & Log-Probabilities 06:13 Anatomy of an LM Training Example 06:37 Computing LLM Probabilities (+Phoebe the Dog) 09:52 Three Common Masks in Post-Training 11:03 A Small Decoding Review 12:14 Training an LM: Cross-Entropy 13:23 Optimization & Fine-Tuning 13:55 Pretraining to Midtraining to SFT Pipeline 15:25 Probability Essentials: KL Divergence & Entropy 19:36 Sigmoid & Pairwise Likelihood 20:29 Reinforcement Learning Framing (MDP) 22:28 Transitioning Tools into Post-Training 23:12 Recommended Resources & Wrap-Up  Happy learning and I'm still taking questions from during the course for Q&A videos.

Nathan Lambert

52,668 次观看 • 4 个月前

I had an awesome time interviewing idan shenfeld and Jonas Hübotter from MIT and ETH Zurich about self-distillation. this very promising post-training paradigm where the model acts as its own teacher by conditioning on environment feedback or demonstrations. we cover the SDPO algo for reinforcement learning with rich feedback and SDFT for continual learning without forgetting along with many applications. we dig into how it works, why it's simpler and faster than GRPO, and where this is already showing up in production systems. table of content: 0:00 - what is self distillation 2:50 - idan (MIT) and jonas (ETH Zurich) introduction and motivation 18:40 - different perspective of on-policy self-distillation (presentation) 36:00 - metacognition and specificity in self-distillation 37:24 - very long hard task and self-distillation 42:00 - continual learning with self-distillation (presentation) 1:16:50 - what is next in this research direction? 1:20:00 - is there any experience with subjective feedbacks? 1:22:50 - quality vs number of feedbacks? 1:26:40 - what setting would self-distillation struggle vs GRPO? the slides were super crisp really cool of them to share! enjoy my guys 🌹
1:31:19

I had an awesome time interviewing idan shenfeld and Jonas Hübotter from MIT and ETH Zurich about self-distillation. this very promising post-training paradigm where the model acts as its own teacher by conditioning on environment feedback or demonstrations. we cover the SDPO algo for reinforcement learning with rich feedback and SDFT for continual learning without forgetting along with many applications. we dig into how it works, why it's simpler and faster than GRPO, and where this is already showing up in production systems. table of content: 0:00 - what is self distillation 2:50 - idan (MIT) and jonas (ETH Zurich) introduction and motivation 18:40 - different perspective of on-policy self-distillation (presentation) 36:00 - metacognition and specificity in self-distillation 37:24 - very long hard task and self-distillation 42:00 - continual learning with self-distillation (presentation) 1:16:50 - what is next in this research direction? 1:20:00 - is there any experience with subjective feedbacks? 1:22:50 - quality vs number of feedbacks? 1:26:40 - what setting would self-distillation struggle vs GRPO? the slides were super crisp really cool of them to share! enjoy my guys 🌹

Yacine Mahdid

13,119 次观看 • 6 个月前

Why isn't the West defending itself? 00:00:00 No Chance of It Being Prime 00:05:24 Sponsor: CrowdHealth 00:07:10 Sponsor: Mars Men 00:11:24 Sponsor: Branch Basics 00:13:57 Kicking Ass and Taking Names with Maps 00:21:24 The UN Adopts Equal Earth 00:26:25 Every Home Should Have a Globe 00:34:03 Is the West at War? 00:37:57 Abortion and Where to Draw the Line 00:46:34 The Lobby and the End of the Republic 00:50:49 Advertising and the Orwell Department 00:53:33 Informed Consent and Fake Blood 01:01:34 Secularism, Utopianism, Liberalism 01:08:39 Is Racism Necessary? 01:13:58 Racism vs. Xenophobia 01:17:51 Colorblind and Not Allowed to Notice 01:23:13 The Danger of the Drifter 01:26:46 Mama Bear Energy and Oxytocin 01:29:27 Cosmopolitan vs. Multicultural 01:35:14 An Honest Discussion About Migration
1:38:20

Why isn't the West defending itself? 00:00:00 No Chance of It Being Prime 00:05:24 Sponsor: CrowdHealth 00:07:10 Sponsor: Mars Men 00:11:24 Sponsor: Branch Basics 00:13:57 Kicking Ass and Taking Names with Maps 00:21:24 The UN Adopts Equal Earth 00:26:25 Every Home Should Have a Globe 00:34:03 Is the West at War? 00:37:57 Abortion and Where to Draw the Line 00:46:34 The Lobby and the End of the Republic 00:50:49 Advertising and the Orwell Department 00:53:33 Informed Consent and Fake Blood 01:01:34 Secularism, Utopianism, Liberalism 01:08:39 Is Racism Necessary? 01:13:58 Racism vs. Xenophobia 01:17:51 Colorblind and Not Allowed to Notice 01:23:13 The Danger of the Drifter 01:26:46 Mama Bear Energy and Oxytocin 01:29:27 Cosmopolitan vs. Multicultural 01:35:14 An Honest Discussion About Migration

Bret Weinstein

30,089 次观看 • 22 天前

The Legitimacy of the American Empire (00:00) The Shifting Definitions of America First (00:24) Realism in American Foreign Policy (02:06) Venezuela: A Case Study (04:13) The American and British Empires (07:05) The Role of Soft Power (14:47) Comparing Modern Empires: Russia and China (17:57) Globalism and the Shadow Empire (23:12) The Legitimacy Crisis of Israel and American Power
26:29

The Legitimacy of the American Empire (00:00) The Shifting Definitions of America First (00:24) Realism in American Foreign Policy (02:06) Venezuela: A Case Study (04:13) The American and British Empires (07:05) The Role of Soft Power (14:47) Comparing Modern Empires: Russia and China (17:57) Globalism and the Shadow Empire (23:12) The Legitimacy Crisis of Israel and American Power

Richard Spencer 🇺🇦

26,926 次观看 • 1 年前

The Legitimacy of the American Empire (00:00) The Shifting Definitions of America First (00:24) Realism in American Foreign Policy (02:06) Venezuela: A Case Study (04:13) The American and British Empires (07:05) The Role of Soft Power (14:47) Comparing Modern Empires: Russia and China (17:57) Globalism and the Shadow Empire (23:12) The Legitimacy Crisis of Israel and American Power
26:29

The Legitimacy of the American Empire (00:00) The Shifting Definitions of America First (00:24) Realism in American Foreign Policy (02:06) Venezuela: A Case Study (04:13) The American and British Empires (07:05) The Role of Soft Power (14:47) Comparing Modern Empires: Russia and China (17:57) Globalism and the Shadow Empire (23:12) The Legitimacy Crisis of Israel and American Power

Richard Spencer 🇺🇦

25,699 次观看 • 1 年前