Internship Experience
Research internships spanning multimodal models, recommendation, and computational advertising.
Alibaba · Taobao & Tmall Group
2025.09 - 2026.05Research Intern · Display Advertising, Ranking Team
- Built MAC, the first public multi-attribution CVR benchmark, integrating First-Click, Last-Click, Linear, and DDA labels. Systematically validated multi-attribution learning across all four attribution objectives, with greater gains for users with complex conversion paths. Online A/B tests achieved GMV +2.7%, orders +1.2%, and ROI +2.6% over the production baseline.
- Proposed LLM-based conversion path understanding to generate semantic attribution signals, identify implicitly related touchpoints missed by rule-based attribution, and incorporate them into auxiliary CVR training. Offline GAUC on the primary conversion task improved by 0.35% over the same-store touchpoint production baseline and by a further 0.15% over the heuristic CABA baseline.
- Produced two first-author papers at KDD 2026 and CIKM 2026.
Conversion AttributionCVR PredictionLarge Language Models
Meituan · LongCat
2026.08 - 2027.01 (expected)Research Intern · Foundation Model Team
- Contributing to LongCat 2.5 and researching Vision Coding in the foundation model team, Basic R&D Platform, Core Local Commerce.
- Independently developing Muse2Web: Benchmarking Coding Agents on Turning Visual Inspiration into Working Websites (ongoing), a benchmark evaluating coding agents' ability to turn visual inspiration into working websites.
Multimodal ModelsVision Coding