4 papers · 1 filter
UniMoMo: Expert Merging-Based MoE Acceleration for Large Recommendation Models
Lei Xin, Bin Gu, Peize Li +8
Sparse mixture-of-experts (MoE) layers expand recommendation capacity through conditional computation, yet a trained checkpoint still stores and routes over its full expert bank. W…
Think with Structured Grounding: Perceptual Reinforcement Learning for Chart and Visual-Tabular Understanding
Changjiang Jiang, Qiannian Zhao, Lei Xin +3
Multimodal Large Language Models (MLLMs) capable of thinking with images often rely on external tools for fine-grained perception. However, this reliance introduces significant inf…
Beyond Human Annotation: Recent Advances in Data Generation Methods for Document Intelligence
Dehao Ying, Fengchang Yu, Haihua Chen +3
The advancement of Document Intelligence (DI) demands large-scale, high-quality training data, yet manual annotation remains a critical bottleneck. While data generation methods ar…
TabDSR: Decompose, Sanitize, and Reason for Complex Numerical Reasoning in Tabular Data
Changjiang Jiang, Fengchang Yu, Haihua Chen +2
Complex reasoning over tabular data is crucial in real-world data analysis, yet large language models (LLMs) often underperform due to complex queries, noisy data, and limited nume…