[updated] - June Liu And Zia

Why Zia sticks with you:

Their 2023 zine “Half-Light / Full Noise” paired June’s photography with Zia’s hand-drawn overlays—ghostly doodles drifting across real city streets. One image of a sleeping bus passenger, with Zia’s stars and thought-bubbles floating above, still lives rent-free in my head. june liu and zia

As Large Language Models (LLMs) are increasingly deployed as mediators in cross-cultural business and diplomatic negotiations, concerns regarding "cultural hallucinations"—confident but incorrect assertions about social norms—have risen. Current models tend to rely on stereotypical averages of cultural values, lacking the nuance to distinguish between high-context professional settings and low-context casual interactions. This paper introduces the , a retrieval-augmented generation approach that dynamically adjusts the 'cultural weight' of model outputs based on user intent and regional specificity. Through a randomized controlled trial involving 240 participants from East Asia and North America, we demonstrate that CAF reduces misunderstandings in simulated negotiations by 34% compared to baseline GPT-4 models. We conclude that explicit, context-aware cultural modeling is essential for the trustworthy deployment of AI in high-stakes communication. Why Zia sticks with you: Their 2023 zine

Once I have more context, I can generate a sample text for you. Current models tend to rely on stereotypical averages

For example, are they: