🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

1/ 🌐Google AI's BARD just dropped a bombshell with its latest YouTube extension feature! But with great power comes great responsibility. Digging deeper into the potential risks and vulnerabilities this update brings. #ai #infosecurity #Security

I tried 14 of the multimodal reasoning examples from the Google DeepMind Gemini paper on OpenAI's chatGPT-4 (with vision). didn't even transcribe the prompts, I just pasted the images of prompts. GPT-4 gets ~12/14 right. 14 part boring thread.

Fun fact I learned today: Meta tried to send GitHub a DMCA takedown when a user uploaded the weights associated with LLaMA. 🦙 How do you think it turned out?

I'm speaking at #pepr24 along with my co-authors Yash Maurya and Vy Tran in Santa Clara, June 3–4, 2024. Discover how LLMs challenge privacy tech! Join me! bit.ly/pepr2024

3/ Inspired by the question, I decided to investigate the 'enum' keyword allowed in function-calling. And guess what? I successfully cracked it! Here's a glimpse of how the jail-break works:

🚨 Are you using empirical benchmarks to evaluate your LLM unlearning method? Our new paper arxiv.org/pdf/2410.02879 investigates how success on these benchmarks can be misleading. A🧵: 1/n

🚨New Paper!🚨 We trained reasoning LLMs to reason about what they don't know. o1-style reasoning training improves accuracy but produces overconfident models that hallucinate more. Meet RLCR: a simple RL method that trains LLMs to reason and reflect on their uncertainty --