Representative Language Generation
This work addresses bias and diversity issues in generative models for AI applications, providing a theoretical foundation, though it is incremental as it builds on prior frameworks.
The paper tackles the problem of diversity and bias in generative models by introducing 'representative generation,' which requires outputs to proportionally represent groups from training data, and it demonstrates feasibility for infinite hypothesis classes under certain conditions but proves negative results for computability with membership queries.
We introduce "representative generation," extending the theoretical framework for generation proposed by Kleinberg et al. (2024) and formalized by Li et al. (2024), to additionally address diversity and bias concerns in generative models. Our notion requires outputs of a generative model to proportionally represent groups of interest from the training data. We characterize representative uniform and non-uniform generation, introducing the "group closure dimension" as a key combinatorial quantity. For representative generation in the limit, we analyze both information-theoretic and computational aspects, demonstrating feasibility for countably infinite hypothesis classes and collections of groups under certain conditions, but proving a negative result for computability using only membership queries. This contrasts with Kleinberg et al.'s (2024) positive results for standard generation in the limit. Our findings provide a rigorous foundation for developing more diverse and representative generative models.