Han Li
7 papers · Latest:
Break the Inaccessible Boundary: Distilling Post-Conversion Content for User Retention Modeling
OCARM uses a two-stage distillation framework to leverage post-conversion content for improved user retention prediction in real-time bidding without feature leakage.
Action-Aware Generative Sequence Modeling for Short Video Recommendation
A2Gen improves short video recommendations by modeling user actions as temporal sequences, leading to significant engagement boosts.
From Local Indices to Global Identifiers: Generative Reranking for Recommender Systems via Global Action Space
GloRank introduces a generative reranking framework for recommender systems that uses global item identifiers instead of local indices, improving item understanding and performance.
Kwai Summary Attention Technical Report
Kwai Summary Attention (KSA) reduces LLM long-context modeling costs by compressing historical contexts into learnable summary tokens.
WebCompass: Towards Multimodal Web Coding Evaluation for Code Language Models
WebCompass is a new multimodal benchmark for evaluating large language models' end-to-end web coding capabilities across generation, editing, and repair tasks.
On the Equivalence Between Auto-Regressive Next Token Prediction and Full-Item-Vocabulary Maximum Likelihood Estimation in Generative Recommendation--A Short Note
This paper proves that auto-regressive next-token prediction in generative recommendation is mathematically equivalent to full-item-vocabulary maximum likelihood estimation.
CodeTracer: Towards Traceable Agent States
CodeTracer helps debug complex code agents by tracing full state transitions and localizing hidden error chains, improving reliability.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.