Abstract People don’t just talk with natural language: sometimes, they talk about it. Dictionaries, language learning resources, scholarly works in linguistics and literature, and even cultural and legal discourse are full of metalanguage explicating linguistic patterns and practices.[…]
Abstract This talk presents a unified perspective on “test-time thinking” as a catalyst for improving generative AI agents through finer-grained reward modeling, data-centric reasoning, and robust alignment. We begin with GenARM, introducing an inductive bias[…]
Abstract I’ll try to make the case for humanities research that is experimental in the scientific sense, involving the systematic manipulation of variables under controlled conditions to support causal inference. To that end, I’ll present[…]
Abstract Generative AI can produce coherent, human-sounding text and arguments. This talk will explore how this capability 1) impacts existing social processes like peer reviewing and sociotechnical infrastructure like crowdsourcing platforms, and 2) may destabilize the[…]
Abstract We propose Universal Speech Content Factorization (USCF), a simple and invertible linear method for extracting a low-rank speech representation in which speaker timbre is suppressed while phonetic content is preserved. USCF extends Speech Content[…]
Abstract Expressive speech generation aims to produce speech that conveys not only linguistic content but also nuanced emotional and stylistic information. However, evaluating the expressiveness of the generated speech remains a challenging problem, often relying[…]
Abstract Natural language processing research has evolved over the past few years from mainly task-specific models, then to task-independent representation models fine-tuned for various tasks, and finally to fully task-independent language models. This progression addresses[…]
