Arman Cohan

This study evaluates when expansions improve or harm information retrieval performance.

This study examines LLM capabilities in producing structured data accurately.

An in-depth look at how LLMs convert language into code across multiple tasks.

A new open language model for research and innovation in natural language processing.

Examining vulnerabilities and safety strategies for LLM-powered scientific agents.

Study reveals significant data overlap affecting language model evaluations in code generation.

A new dataset helps IR models adapt to complex instructions for better performance.

Data contamination affects the evaluation of large language models significantly.

Two methods enhance the accuracy of AI-generated text evaluations.

A new benchmark assesses models for verifying financial claims in complex documents.

ChemSafetyBench tests chatbots on chemical safety and knowledge.