Sean Hendryx
@SeanHendryx
AI research & engineering at Scale
What will the learning environments of the future look like that train artificial super intelligence? In recent work at @scale_AI , we show that training systems that combine verifiable rewards with multi-agent interaction accelerate learning.
New @Scale_AI paper! 🌟 LLMs trained with RL can exploit reward hacks but not mention this in their CoT. We introduce verbalization fine-tuning (VFT)—teaching models to say when they're reward hacking—dramatically reducing the rate of undetected hacks (6% vs. baseline of 88%).
🤔 How do we train LLMs on real-world tasks where it’s hard to define a single verifiable answer? Our work at @scale_AI introduces Rubrics as Rewards (RaR) — a framework for on-policy post-training that uses structured, checklist-style rubrics as interpretable reward signals. 🧵
We’re entering a new era in robotics where generalized systems are starting to work in the real world, but researchers still don’t have good tools for understanding their data. That’s why I built ARES, an open-source platform for ingesting, annotating, and curating robotics data.
🤖 AI agents are crossing into the real world. But when they act independently—who’s watching? At Scale, we’re building Agent Oversight: a platform to monitor, intervene, and align autonomous AI. We’re hiring engineers (SF/NYC) to tackle one of the most urgent problems in AI.…
If a model lies when pressured—it’s not ready for AGI. The new MASK leaderboard is live. Built on the private split of our open-source honesty benchmark (w/ @ai_risks), it tests whether models lie under pressure—even when they know better. 📊 Leaderboard:…
Do safety-aligned LLMs remain safe when used as browser agents? Our paper "Refusal-Trained LLMs Are Easily Jailbroken As Browser Agents", accepted to ICLR, shows they don't. This reveals a critical gap in current alignment approaches.
“If basic research is so valuable then the market will fund it” is an argument that is directly addressed in like AP economics. When you make that argument without acknowledging its standard rebuttal, you reveal that you stopped paying attention after the fifth lecture.