Former ByteDance intern wins NeurIPS award amid legal battle
Former intern Tian Keyu's paper, co-authored with ByteDance, was named Best Paper at top-level AI conference NeurIPS, sparking widespread discussion.
The recognition was notable not only due to the influence of NeurIPS (Neural Information Processing Systems) in the industry but also because ByteDance, TikTok's parent firm, recently sought legal action against Tian, demanding eight million yuan (US$1.1 million) for "poisoning" algorithms.
The Best Paper certificate with Tian's name as first author.
NeurIPS is an international conference focused on machine learning and computational neuroscience. Normally held in December each year, it is ranked as a Class A conference in the field of AI.
The award-winning paper, co-published with Peking University, titled "Visual Autoregressive Modeling: Scalable Image Generation through Next-Scale Prediction," explores a scalable method for image generation using Visual Autoregressive Modeling (VAR), with Tian listed as the first author.
The conflict between ByteDance and Tian escalated to a lawsuit after allegations that Tian maliciously disrupted or was "poisoning" internal model training by altering code, leading to significant resource wastage.
ByteDance has clarified that the incident only affected a research project within its commercial technology team and did not impact formal projects or other businesses, dismissing online claims of "tens of millions of dollars in losses" as gross exaggerations.
The legal dispute continues as ByteDance pursues compensation and a public apology from Tian, highlighting the complexities that can arise from the intersection of academic achievements and corporate governance.
