Dawn Song
@dawnsongtweets
Professor in Computer Science at UC Berkeley; Building safe, responsible, decentralized AI; Serial entrepreneur
1/ 🔥 AI agents are reaching a breakthrough moment in cybersecurity. In our latest work: 🔓 CyberGym: AI agents discovered 15 zero-days in major open-source projects 💰 BountyBench: AI agents solved real-world bug bounty tasks worth tens of thousands of dollars 🤖…
Really excited to share our recent work: “VERINA: Benchmarking Verifiable Code Generation.” As more code underpinning our digital world is generated by AI, we need to move beyond asking “does it run?” to “can we trust that it’s correct and secure?” VERINA is a step toward that…
1/🧵Introducing VERINA: a high-quality benchmark for verifiable code generation. As LLMs are increasingly used to generate software, we need more than just working code--We need formal guarantees of correctness. VERINA offers a rigorous and modular framework for evaluating LLMs…
My group & collaborators have developed many popular benchmarks over the years, e.g., MMLU, MATH, APPS---really excited about our latest benchmark OMEGA Ω: 🔍Can LLMs really think outside the box in math? a new benchmark probing 3 axes of generalization: 1️⃣ Exploratory 2️⃣…
📢 Can LLMs really reason outside the box in math? Or are they just remixing familiar strategies? Remember DeepSeek R1, o1 have impressed us on Olympiad-level math but also they were failing at simple arithmetic 😬 We built a benchmark to find out → OMEGA Ω 📐 💥 We found…
Really excited about our work on "Learning to Reason without External Rewards", proposing a new paradigm: Reinforcement Learning from Internal Feedback (RLIF)--- Without access to ground-truth answers, simply by optimizing their own internal sense of confidence, LLMs can learn…
🚀 Excited to share the most inspiring work I’ve been part of this year: "Learning to Reason without External Rewards" TL;DR: We show that LLMs can learn complex reasoning without access to ground-truth answers, simply by optimizing their own internal sense of confidence. 1/n