Christopher Potts

Full Professor@Stanford University·美国·OpenReview

研究方向

Linguistics · Perception · Action & Cognition · Robust Intelligence · Action and Cognition · Interpretability · Explainable AI · Information Retrieval

Christopher Potts

GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning

Addressing divergent representations from causal interventions on neural networks

Diagnosing Bottlenecks in Data Visualization Understanding by Vision-Language Models

Mechanistic evaluation of Transformers and state space models

Multi-module GRPO: Composing Policy Gradients and Prompt Optimization for Language Model Programs

Improved Representation Steering for Language Models

AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders

Blackbox Model Provenance via Palimpsestic Membership Inference

Internal Causal Mechanisms Robustly Predict Language Model Out-of-Distribution Behaviors

Do Language Models Use Their Depth Efficiently?

HyperDAS: Towards Automating Mechanistic Interpretability with Hypernetworks

Base Models Beat Aligned Models at Randomness and Creativity

Bayesian scaling laws for in-context learning

Improving Pretraining Data Using Perplexity Correlations

Bayesian scaling laws for in-context learning

MrT5: Dynamic Token Merging for Efficient Byte-level Language Models