AI.info

Thomas Foster

Explore Thomas Foster on AI.info.

• LLMs Encode Their Failures: Predicting Success from Pre-Generation Activations
• Measuring what Matters: Construct Validity in Large Language Model Benchmarks