Verified facts
What happened
The paper, titled MERIT: Mitigating Exposure Bias in Generative XMC for User-Interest Propensity Modeling, was published in the 2026 record as a preprint.
The listed authors are Abhinav Mahajan, Arindam Sarkar, and Prakash Mandayam Comar.
The source is the arXiv cs.AI Daily Feed, with identifier 2608.28931.
The available evidence is bibliographic metadata only; no abstract, full text, quotation, method, sample, result, or effect size is available.
Business relevance
Why it matters
User-interest propensity models can influence audience segmentation, recommendation, search ranking, and product discovery, so exposure bias is a relevant issue for merchants using algorithmic targeting.
The title points to a research concern around whether observed user interests reflect genuine preferences or simply what users were previously shown, but the supplied record does not establish that