David Brandfonbrener

Researcher@Anthropic·美国·OpenReview

研究方向

large language models · representation learning · statistical learning theory · generalization · reinforcement learning · offline · optimization · exploration · regret bounds

7.5

The Art of Scaling Reinforcement Learning Compute for LLMs

ICLR 2026Oral

5.5

Generalized Parallel Scaling with Interdependent Generations

ICLR 2026Poster

二作

4.0

Let's (not) just put things in Context: Test-time Training for Long-context LLMs

合作者 (20)

David Brandfonbrener

The Art of Scaling Reinforcement Learning Compute for LLMs

Generalized Parallel Scaling with Interdependent Generations

Let's (not) just put things in Context: Test-time Training for Long-context LLMs

Mixture of Parrots: Experts improve memorization more than reasoning

The Role of Sparsity for Length Generalization in LLMs

SOAP: Improving and Stabilizing Shampoo using Adam for Language Modeling

Deconstructing What Makes a Good Optimizer for Autoregressive Language Models

Universal Length Generalization with Turing Programs

Universal length generalization with Turing Programs