Modeling Human-Like Color Naming Behavior in Context
Yuqing Zhang, Ecesu Ürker, Tessa Verhoef, Gemma Boleda, Arianna Bisazza
TLDR
This paper improves neural agent models for color naming by using upsampling and multi-listener interactions to generate more human-like, convex color categories.
Key contributions
- Identified that neural agent models for color naming produce non-convex categories, unlike human systems.
- Introduced upsampling rare color terms and multi-listener RL to improve human-like color naming.
- Showed upsampling enhances lexical diversity, while multi-listener setups promote convex color categories.
- Found combining both methods yields color lexicons most similar to human systems.
Why it matters
This paper addresses a key limitation in computational models of color naming, where generated lexicons diverge from human perception. By introducing novel training factors, it significantly improves the human-likeness of these models. This advancement is crucial for developing AI systems that can better understand and interact with human language and cognitive processes.
Original Abstract
Modeling the emergence of human-like lexicons in computational systems has advanced through the use of interacting neural agents, which simulate both learning and communicative pressures. The NeLLCom-Lex framework (Zhang et al., 2025) allows neural agents to develop pragmatic color naming behavior and human-like lexicons through supervised learning (SL) from human data and reinforcement learning (RL) in referential games. Despite these successes, the lexicons that emerge diverge systematically from human color categories, producing highly non-convex regions in color space, which contrast with the convexity typical of human categories. To address this, we introduce two factors, upsampling rare color terms during SL and multi-listener RL interactions, and adopt a convexity measure to quantify geometric coherence. We find that upsampling improves lexical diversity and system-level informativeness of the color lexicon, while many-listener setups promote more convex color categories. The combination of moderate upsampling and multiple listeners produces lexicons most similar to human systems.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.