5 papers
Post-Training is About States, Not Tokens: A State Distribution View of SFT, RL, and On-Policy Distillation
Dong Nie
Large language model post-training methods such as supervised fine-tuning (SFT), reinforcement learning (RL), and distillation are often analyzed through their loss functions: maxi…
Sub-JEPA: Subspace Gaussian Regularization for Stable End-to-End World Models
Kai Zhao, Dongliang Nie, Yuchen Lin +4
Joint-Embedding Predictive Architectures (JEPAs) provide a simpleframework for learning world models by predicting future latent representations.However, JEPA training is subject t…
UNA: A Unified Supervised Framework for Efficient LLM Alignment Across Feedback Types
Zhichao Wang, Bin Bi, Can Huang +7
RL alignment methods, including RLHF and DPO, are primarily based on pairwise preference data. Although scalar or score-based feedback has been collected in some settings, it is ra…
Seeing the Forest and the Trees: Query-Aware Tokenizer for Long-Video Multimodal Language Models
Siyou Li, Huanan Wu, Juexi Shao +10
Despite the recent advances in the video understanding ability of multimodal large language models (MLLMs), long video understanding remains a challenge. One of the main issues is…
Tokenizer: Differentiable Multi-Scale Multi-Modal Tokenizer for Radiology Report Generation
Siyou Li, Pengyao Qin, Huanan Wu +4
Automated radiology report generation (RRG) aims to produce detailed textual reports from clinical imaging, such as computed tomography (CT) scans, to improve the accuracy and effi…