After almost 3 years, I have finally learned the basics of the hottest programming language. Okay, now what's the next step to pursue a master's degree?
رَبُّكُمْ أَعْلَمُ بِمَا فِي نُفُوسِكُمْ ۚ إِنْ تَكُونُوا صَالِحِينَ فَإِنَّهُ كَانَ لِلْأَوَّابِينَ غَفُورًا
This paper asks when LLMs can be trusted to judge mental health replies. Found that LLMs systematically overrate replies, especially on empathy and helpfulness. Even when the ranking order matched human experts, the actual scores were too high, which means models look better…
مش عارفه كنت عايشة ازاي من غير ما اذاكر System Design كل ما الواحد يعرف اكتر في ال software engineering يحس ان صناعة ال software ممتعة ومعقدة مش شوية كود الواحد يكتبهم ويلزقهم مع بعض
🔥 GPT-6 may not just be smarter, it might be alive (in the computational sense). A new research paper called SEAL, Self-Adapting Language Models (arXiv:2506.10943) describes how an AI can continuously learn after deployment, evolving its own internal representations without…
Check out their blogs if you are into AI/ML. 1) Andrej Karpathy Neural networks & LLMs explained from first principles by one of the OGs of modern AI. - karpathy.ai/?utm_source=li… 2) Sebastian Raschka, PhD Deep dives into LLM training and fine-tuning with super clear code…
RIP fine-tuning ☠️ This new Stanford paper just killed it. It’s called 'Agentic Context Engineering (ACE)' and it proves you can make models smarter without touching a single weight. Instead of retraining, ACE evolves the context itself. The model writes, reflects, and edits…
- build an autograd engine from scratch - write a mini-GPT from scratch - implement LoRA and fine-tune a model on real data - hate CUDA at least once - cry - keep going the roadmap - 5 phases - if you already know something? skip - if you're lost? rewatch - if you’re stuck? use…
PPO vs GRPO vs REINFORCE – a workflow breakdown of the most talked-about reinforcement learning algorithms ➡️ Proximal Policy Optimization (PPO): The Stable Learner It’s used everywhere from dialogue agents to instruction tuning as it balances between learning fast and staying…
Ironically, LLM researchers know nothing about language.
This one paper might kill the AI scaling hype. While Big Tech burns billions on massive datasets, researchers just achieved state-of-the-art agent performance using 78 samples. And it makes a scary amount of sense. Here's the full breakdown:
the closer I get to my MVP being done, the more I don’t want to ship it lol
بفتكر ايام ثانوية عامة ( و اول سنتين فالكلية) لما كان جزء من المدرسين و معيدين الكلية يعاملوني معاملة غريبة فشخ و لما اسألهم يقولولي "هو كدا شكلك مش عاجبني" لحد ما قررت استلسم و احلق
نائب رئيس الجامعه و هو بيسلمني شهاده تقدير عشان اتأهلنا لل ACPC من يومين بدل ما يقولي مبروك بيقولي ينفع شعرك دا و انا اتقتلت ضحك من بعدها قدامهم و حتي الصوره الي منزلنها باين فيها اني كنت مقتول ضحك
حاجة كدا زي "عشان تعرف حسابك قوي ولا لا اكتب الباسورد في كومنت لو ظهر **** يبقي حسابك قوي"
This Tencent paper shows a way to improve reasoning by training only on raw text using reinforcement learning. It is called Reinforcement Learning on Pre-Training data (RLPT) and it removes the need for human labels. Simple “predict the next segment” rewards are enough to…
"Rule number one, DON'T FKN SAY 'you are absolutely right!' and write code!"
DSPy and ColBERT are interesting academic experiments imo. Each is a multi-paper repo that has one coherent artifact, combining our latest research together. We typically release the features as open source—hence get users/feedback—well before writing a paper on the new ideas.
Oh yes, I meant that I thought a paper came out with the thread you made about DSPy in August.
Talking to grad students, too many think that long-term projects (not scattered papers), proper code releases, thoughtful benchmarks are "not incentivized". Most often they're mistaken. If we're talking incentives, *nothing* matches demonstrating impact! Will blog on this soon.
Science of RL optimization is likely humanity’s last open scientific problem
Prediction: In ~3 years academia will be the most desirable place to do fundamental AI research Contributing factors: - small models improve/become significantly more impactful - open weights community broadens its reach - gpus continue to get faster & cheaper - meaningful…
United States 趨勢
- 1. Good Saturday 20.6K posts
- 2. Tosin 6,273 posts
- 3. #SaturdayVibes 3,051 posts
- 4. #LingOrm3rdMeetMacauD1 474K posts
- 5. LINGORM MACAU MEET D1 471K posts
- 6. Travis Head 25.8K posts
- 7. Burnley 23.8K posts
- 8. Marjorie Taylor Greene 81.3K posts
- 9. #MyHeroAcademia 16.4K posts
- 10. The View 98.6K posts
- 11. Somali 79.3K posts
- 12. Caleb Love 3,319 posts
- 13. #GoDeku 9,205 posts
- 14. #DBLF2025 17.7K posts
- 15. Kandi 12.2K posts
- 16. Collin Gillespie 4,367 posts
- 17. Suns 23.2K posts
- 18. Norvell 3,685 posts
- 19. Joshua 109K posts
- 20. Sengun 8,898 posts
Something went wrong.
Something went wrong.