Full video: 🎥 https://youtu.be/UwX5zzjwb_g Large language models often memorize what they see — even a single phone number or address can stick forever in their weights. Google’s new VaultGemma changes that: it’s the first open-weight LLM trained from scratch with differential privacy, meaning secrets seen seldomly during training leaves no trace. 👉 In this video, we explain Differential Privacy through the concrete example of VaultGemma — how it works, why it matters, and what it means for the future of trustworthy AI. AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Vignesh Valliappan, Ivan Janov, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

LLMs often memorize what they see — even a single phone number or address can stick forever in their weights. Google’s new VaultGemma changes that: it’s the first open-weight LLM trained from scratch with differential privacy, meaning secrets seen seldomly during training leaves no trace. 👉 In this video, we explain Differential Privacy through the concrete example of VaultGemma — how it works, why it matters, and what it means for the future of trustworthy AI. AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Vignesh Valliappan, Ivan Janov, Sunny Dhiana, Andy Ma 📃 VaultGemma Blog: https://research.google/blog/vaultgemma-the-worlds-most-capable-differentially-private-llm/ 📃VaultGemma Paper: https://services.google.com/fh/files/blogs/vaultgemma_tech_report.pdf Outline: 00:00 VaultGemma explained 00:58 Differential Privacy explained 02:33 Training with Differential Privacy 05:22 Training recap 06:09 Results 07:16 Real-world impact ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

Full video: https://youtu.be/firXjwZ_6KI AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ We explain diffusion models and flow-matching models side by side to highlight the key differences between them. Flow-Matching Models are the new generation of AI image generators that are quickly replacing diffusion models — they take everything diffusion did well, but make it faster, smoother, and deterministic. Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Vignesh Valliappan, Ivan Janov, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

We explain diffusion models and flow-matching models side by side to highlight the key differences between them. Flow-Matching Models are the new generation of AI image generators that are quickly replacing diffusion models — they take everything diffusion did well, but make it faster, smoother, and deterministic. AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ Text to image diffusion models: https://youtu.be/J87hffSMB60 Useful deeper reading: • 🌊 Lipman et al., “Flow Matching for Generative Modeling” (2023) — https://arxiv.org/abs/2210.02747 • 🧮 Kingma and Gao, "Understanding Diffusion Objectives as the ELBO with Simple Data Augmentation" (2022) — https://arxiv.org/abs/2210.02747 • ⚡ Esser et al, "Scaling Rectified Flow Transformers for High-Resolution Image Synthesis" (2024) — https://arxiv.org/abs/2403.03206 Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Vignesh Valliappan, Ivan Janov, Sunny Dhiana, Andy Ma Outline: 00:00 Difference between Flow-matching and Diffusion 01:07 Training Diffusion Models 05:45 Inference for Diffusion Models 09:03 Training Flow-Matching 11:55 Inference with Flow-Matching 14:02 Side-by-Side Comparison ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

full video ► https://youtu.be/18Fn2m99X1k Energy Based Transformers and Models explained AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Vignesh Valliappan, Ivan Janov, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

I always wanted to know how energy-based models (EBMs) work. In this video, we break down EBMs — what they are, how they work, and how they’re different from standard neural networks. ► Then we zoom in on the Energy-Based Transformers (EBTs) paper by Gladstone et al. 2025, showing how the authors combined EBMs with transformers to create models that can refine their guesses, self-verify, and potentially adapt how much computation they use depending on the difficulty of the problem. AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Vignesh Valliappan, Ivan Janov, Sunny Dhiana, Andy Ma 📃 Alexi Gladstone, Ganesh Nanduru, Md Mofijul Islam, Peixuan Han, Hyeonjeong Ha, Aman Chadha, Yilun Du, Heng Ji, Jundong Li, and Tariq Iqbal. "Energy-Based Transformers are Scalable Learners and Thinkers." (2025 https://arxiv.org/abs/2507.02092 ) Outline: 00:00 Energy-Based Transformers 00:47 EBT paper 01:07 Energy-based models explained 03:56 EBM training 06:37 EBM inference 07:55 Energy-based LLM 09:07 Stabilising Training via Noise 09:59 Replay Buffer 10:34 Randomising Step Size 10:26 Results 12:54 Energy-based Transformers for other Modalities ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

ACL 2025 just took place in Vienna — the world’s largest NLP conference with almost 2,000 papers presented! 🎓✨ Here is a quick snapshot of the event via a short interview with one of the authors whose work caught my attention. AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ 1️⃣ Fabio j. Fehr @ EPFL & Idiap 📃Fabio James Fehr, Prabhu Teja S, Luca Franceschi, and Giovanni Zappella. 2025. CoRet: Improved Retriever for Code Editing. ACL 2025, Vienna 2️⃣Frederick Riemenschneider @Heidelberg University 📃 Frederick Riemenschneider and Anette Frank. 2025. Cross-Lingual Generalization and Compression: From Language-Specific to Shared Neurons. ACL 2025, Vienna Outline: 00:00 From Probabilities to Words 0:33 CoRet: Improved Retriever for Code Editing by Fabio j. Fehr @ EPFL & Idiap 3:15 Cross-Lingual Generalisation and Compression by Frederick Riemenschneider @Heidelberg University 8:50 Outro from ACL Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Dres. Trost GbR, Siltax, Vignesh Valliappan, Michael, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

How LLMs Choose the Next Word – Decoding Strategies Explained, here we explain why min-p sampling became so popular in today's LLMs. Learn more in the full video: https://youtu.be/o-_SZ_itxeA AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Dres. Trost GbR, Siltax, Vignesh Valliappan, Michael, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

How LLMs Choose the Next Word – Decoding Strategies Explained, here we discuss the effect of temperature in sampling. Learn more in the full video: https://youtu.be/o-_SZ_itxeA AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Dres. Trost GbR, Siltax, Vignesh Valliappan, Michael, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

Did you know AI doesn’t always pick the top word? 🤯 Here’s how Top-p sampling makes text more creative! Learn more in the full video: https://youtu.be/o-_SZ_itxeA AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Dres. Trost GbR, Siltax, Vignesh Valliappan, Michael, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

Let's debunk the lie that LLMs choose the next most probable token AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Dres. Trost GbR, Siltax, Vignesh Valliappan, Michael, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

How LLMs Choose the Next Word – Decoding Strategies Explained Learn more in the full video: https://youtu.be/o-_SZ_itxeA AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Dres. Trost GbR, Siltax, Vignesh Valliappan, Michael, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

How do large language models like ChatGPT actually decide which word comes next? In this video, we break down the core decoding strategies used in text generation: from greedy decoding to top-k, top-p (nucleus sampling), and the newer min-p sampling. You’ll learn how these methods turn probability distributions into actual words, and how just changing the sampling strategy can make the same model sound repetitive, brilliant, or totally unpredictable. We also show hands-on examples using GPT-2 to demonstrate how decoding choices affect output, without changing the model itself. AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ 💻 Collab used in this video: https://drive.google.com/file/d/14OyZRyUE3iVTVK5MG8ZyXReuzZwgg5Xg/view?usp=sharing 📃 Minh, Nguyen Nhat, Andrew Baker, Clement Neo, Allen Roush, Andreas Kirsch, Wand AI Independent, Ravid Shwartz-Ziv, and Wand AI. "Turning Up the Heat: Min-p Sampling for Creative and Coherent LLM Outputs." In The Thirteenth International Conference on Learning Representations. 2024. https://arxiv.org/abs/2407.01082 Outline: 00:00 From Probabilities to Words 00:57 Recap: Next Token Prediction Basics 02:05 Deterministic: Greedy Decoding 02:55 Why Sampling Matters 03:52 Random Decoding 04:56 Top-k Sampling 05:46 Top-p Sampling (Nucleus Sampling) 06:54 Temperature 08:11 Min-p Sampling 09:29 Repetition & Frequency Penalty 09:52 Beam Search 10:43 Summary & Takeaways Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Dres. Trost GbR, Siltax, Vignesh Valliappan, Michael, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost Music 🎵 : Space Navigator – Sarah, The Illstrumentalist

AlphaEvolve does not just write code, but it evolves it into better and better solutions, all on its own. In this video, we explain AlphaEvolve, DeepMind’s latest coding agent that uses large language models and evolutionary strategies to discover new algorithms and optimize real-world systems. From improving matrix multiplication to cutting training time for large models, AlphaEvolve shows what happens when AI becomes a tool in research. AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ 📃 Alexander Novikov, Ngân Vu, Marvin Eisenberger, Emilien Dupont, Po-Sen Huang, Adam Zsolt Wagner, Sergey Shirobokov et al. "AlphaEvolve: A coding agent for scientific and algorithmic discovery." Google DeepMind (2025). https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/AlphaEvolve.pdf Outline: 00:00 AlphaEvolve explained 00:34 In a nutshell 01:01 AlphaEvolve in detail 02:33 Evolutionary strategies: MAP-Elites 03:35 Island population models 04:42 Results 05:58 Discussion 07:08 Connection to FunSearch Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Dres. Trost GbR, Siltax, Vignesh Valliappan, Michael, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost

Long videos are a nightmare for language models—too many tokens to handle, plus many tokens are redundant, slow inference, and limited context windows. STORM ⛈️ changes that. In this AI Coffee Break, we explain STORM, a new architecture from NVIDIA and collaborators that improves long video understanding using Mamba layers for temporal modeling and token compression. The result? Better accuracy than GPT-4o on key benchmarks and up to 8× more efficiency. AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring.com/ ☕ Grab a cup of coffee and learn: • Why current Video LLMs struggle with long sequences • How STORM uses Mamba to inject temporal context before compression • How it reduces visual tokens by a factor of 8—without sacrificing performance • Benchmarks: MVBench, MLVU, and beyond • Why this is a big step toward real-world video comprehension 📃 Jindong Jiang, Xiuyu Li, Zhijian Liu, Muyang Li, Guo Chen, Zhiqi Li, De-An Huang et al. "Token-Efficient Long Video Understanding for Multimodal LLMs." (2025) https://arxiv.org/abs/2503.04130 Outline: 00:00 Long sequence struggle 01:36 Video LLMs so far 02:20 The idea for STORM 03:31 Training details 07:01 Results 08:18 Discussion Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Dres. Trost GbR, Siltax, Vignesh Valliappan, Michael, Sunny Dhiana, Andy Ma ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel as a Bean Member to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter / X: https://twitter.com/AICoffeeBreak LinkedIn: https://www.linkedin.com/in/letitia-parcalabescu/ Threads: https://www.threads.net/@ai.coffee.break Bluesky: https://bsky.app/profile/aicoffeebreak.bsky.social Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak Substack: https://aicoffeebreakwl.substack.com/ Web: https://explanationmark.de/letitia https://aicoffeebreak.com #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost