Nova Dassarma

发表

Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training

Daniel M. Ziegler, D. Duvenaud, P. Christiano, 2024, ArXiv.

Discovering Language Model Behaviors with Model-Written Evaluations

Tom B. Brown, Dario Amodei, T. Henighan, 2022, ACL.

Constitutional AI: Harmlessness from AI Feedback

Tom B. Brown, Dario Amodei, Azalia Mirhoseini, 2022, ArXiv.

Measuring Progress on Scalable Oversight for Large Language Models

Tom B. Brown, Dario Amodei, Azalia Mirhoseini, 2022, ArXiv.

Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned

Tom B. Brown, Dario Amodei, T. Henighan, 2022, ArXiv.

A General Language Assistant as a Laboratory for Alignment pdf

Dario Amodei, Tom Henighan, Tom B. Brown, 2021, ArXiv.

Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback

Tom B. Brown, Dario Amodei, T. Henighan, 2022, ArXiv.

In-context Learning and Induction Heads

Tom B. Brown, Dario Amodei, T. Henighan, 2022, ArXiv.

Scaling Laws and Interpretability of Learning from Repeated Data

Tom B. Brown, Dario Amodei, T. Henighan, 2022, ArXiv.

Language Models (Mostly) Know What They Know

Tom B. Brown, Dario Amodei, T. Henighan, 2022, ArXiv.