Show HN: I put a $2.43 necklace on 3 outfits. VLMs priced it at $19 to $104
Summary
A companion study investigates context-driven valuation bias (halo effect) in six multimodal LLMs, showing that an identical necklace is valued differently depending on outfit/context, with several models revealing biases in image vs. text prompts. The repo provides reproduction steps and analyzes multiple models' responses, highlighting implications for AI-assisted valuation tasks.