Paper · arXiv 2409.00729
ContextCite: Attributing Model Generation to Context
Benjamin Cohen-Wang, Harshay Shah, Kristian Georgiev, Aleksander Madry
ContextCite is a method for attributing parts of the context used by language models to generate statements, aiding in verification, improving response quality, and detecting poisoning attacks.
Abstract
How do language models use information provided as context when generating a response? Can we infer whether a particular generated statement is actually grounded in the context, a misinterpretation, or fabricated? To help answer these questions, we introduce the problem of context attribution: pinpointing the parts of the context (if any) that led a model to generate a particular statement. We then present ContextCite, a simple and scalable method for context attribution that can be applied on top of any existing language model. Finally, we showcase the utility of ContextCite through three applications: (1) helping verify generated statements (2) improving response quality by pruning the context and (3) detecting poisoning attacks. We provide code for ContextCite at https://github.com/MadryLab/context-cite.