5 papers
Synthetic Persona Pretraining: Alignment from Token Zero
Julian Minder, Viktor Moskvoretskii, Raghav Singhal +12
As language-model-based AI is increasingly deployed in autonomous settings, aligning its goals and values with those of humans becomes critical. Today, alignment, and the assistant…
From Model Training to Model Raising
Roland Aydin, Christian Cyron, Steve Bachelor +2
Current AI training methods align models with human values only after their core capabilities have been established, resulting in models that are easily misaligned and lack deep-ro…
Tandem Training for Language Models
Robert West, Ashton Anderson, Ece Kamar +1
As language models continue to rapidly improve, we can expect their actions and reasoning to become difficult or impossible for weaker agents and humans to follow, undermining inte…
Web2Wiki: Characterizing Wikipedia Linking Across the Web
Veniamin Veselovsky, Tiziano Piccardi, Ashton Anderson +2
Wikipedia is one of the most visited websites globally, yet its role beyond its own platform remains largely unexplored. In this paper, we present the first large-scale analysis of…
Capturing Dynamics in Online Public Discourse: A Case Study of Universal Basic Income Discussions on Reddit
Rachel Kim, Veniamin Veselovsky, Ashton Anderson
Societal change is often driven by shifts in public opinion. As citizens evolve in their norms, beliefs, and values, public policies change too. While traditional opinion polling a…