David Alvarez-Melis

Assistant Professor@School of Engineering and Applied Sciences, Harvard University·美国·OpenReview

研究方向

Large Language Models · Optimal Transport · Interpretability · Deep Learning · Neural Networks · Natural Language Processing · Machine Learning

6.0

Boomerang Distillation Enables Zero-Shot Model Size Interpolation

ICLR 2026Poster

通讯

5.0

Adam or Gauss-Newton? — A Comparative Study In Terms of Basis Alignment and SGD Noise

ICLR 2026Rejected

4.0

Guided Speculative Inference for Efficient Test-Time Alignment of LLMs

ICLR 2026Poster

三作

4.0

Can Interpretation Predict Behavior on Unseen Data?

ICLR 2026Withdrawn

4.0

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

合作者 (20)

David Alvarez-Melis

Boomerang Distillation Enables Zero-Shot Model Size Interpolation

Adam or Gauss-Newton? — A Comparative Study In Terms of Basis Alignment and SGD Noise

Guided Speculative Inference for Efficient Test-Time Alignment of LLMs

Can Interpretation Predict Behavior on Unseen Data?

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

Random Scaling of Emergent Capabilities

To Backtrack or Not to Backtrack: When Sequential Search Limits Model Reasoning

Mixture of Parrots: Experts improve memorization more than reasoning

Sometimes I am a Tree: Data Drives Unstable Hierarchical Generalization

ICDA: Interactive Causal Discovery through Large Language Model Agents