Zhouhan Lin

Flow of Spans: Generalizing Language Models to Dynamic Span-Vocabulary via GFlowNets

FlowRL: Matching Reward Distributions for LLM Reasoning

AWM: Accurate Weight-Matrix Fingerprint for Large Language Models

PonderLM: Pretraining Language Models to Ponder in Continuous Space

MLP Memory: A Retriever-Pretrained Memory for Large Language Models

Context-level Language Modeling by Learning Predictive Context Embeddings

Pretraining LLM with Latent Thoughts in Continuous Space

FreqKV: Key-Value Compression in Frequency Domain for Context Window Extension

Towards Compressive and Scalable Recurrent Memory

$\alpha$-GFN: Generalized Mixing in GFlowNets for Better Exploration-Exploitation Trade-off

VQKV: High-Fidelity and High-Ratio Cache Compression via Vector-Quantization

Memory Decoder: A Pretrained, Plug-and-Play Memory for Large Language Models

Token Representation Shrinkage Impairs Creativity of Generative Models

ToEdit: How to Synthesize Text Data to Avoid Model Collapse?

AdaptiveStep: Automatically Dividing Reasoning Step through Model Confidence

Training-free LLM-generated Text Detection by Mining Token Probability Sequences

FreqKV: Frequency Domain Key-Value Compression for Efficient Context Window Extension

How to Synthesize Text Data without Model Collapse?

Gradient based Causal Discovery with Diffusion Model

FARV: Leveraging Facial and Acoustic Representation in Vocoder For Video-to-Speech Synthesis