Researcher Profile

Francis Song

Red teaming with language models

Co-author, Red Teaming LMs

Co-authored Red Teaming LMs with LMs: a concrete approach to stress-testing model behavior at scale.

Topics

Post-Training & Alignment Evaluation & Benchmarks Security & Robustness

About This Page

This profile is meant to help you get oriented quickly: why this researcher matters, what to read first, and where to explore next.

Last updated

March 20, 2026

Best First Clicks

Red Teaming Language Models with Language Modelspaper

Known For

The ideas, systems, and research directions that make this person worth knowing.

Red teaming with language models

Red Teaming Language Models with Language Models

Safety

Red teaming

Red Teaming

Start Here

Canonical papers, project pages, or repositories that anchor this profile.

Red Teaming Language Models with Language Modelspaper

Related Researchers

People worth exploring next because they share topics, labs, or source material with this profile.

Shared canonical source

John Aslanides

Red teaming with language models

1 source

Co-authored Red Teaming LMs with LMs: a concrete approach to stress-testing model behavior at scale.

Post-Training & Alignment Evaluation & Benchmarks

Start HereRed Teaming Language Models with Language Models

Shared canonical source

Nat McAleese

Red teaming with language models

1 source

Co-authored Red Teaming LMs with LMs: a concrete approach to stress-testing model behavior at scale.

Post-Training & Alignment Evaluation & Benchmarks

Start HereRed Teaming Language Models with Language Models

Shared canonical source

Saffron Huang

Red teaming with language models

1 source

Co-authored Red Teaming LMs with LMs: a concrete approach to stress-testing model behavior at scale.

Post-Training & Alignment Evaluation & Benchmarks

Start HereRed Teaming Language Models with Language Models

Shared topics

Nicholas Carlini

Adversarial ML, security of deployed models

4 sources

One of the most useful people to study if you care about what deployed models get wrong under pressure, especially around extraction, adversarial behavior, and practical security failures.

Post-Training & Alignment Evaluation & Benchmarks

Start HereTowards Evaluating the Robustness of Neural Networks

Shared topics

Sandipan Kundu

Alignment via AI feedback (Constitutional AI)

4 sources

Useful for the attack-and-evaluation side of alignment work, especially long-context jailbreak research and the measurement work that turns safety concerns into concrete tests.

Anthropic Post-Training & Alignment Evaluation & Benchmarks

Start HereMany-shot Jailbreaking

Shared topics

Ethan Perez

Alignment via AI feedback (Constitutional AI)

4 sources

Important because he sits near the boundary between alignment theory and concrete failure-mode discovery, especially jailbreaks, preference training, and behavior evaluations.

Anthropic Post-Training & Alignment Evaluation & Benchmarks

Start HereConstitutional AI: Harmlessness from AI Feedback