Xu Yang

CoT Vectors: Transferring and Probing the Reasoning Mechanisms of LLMs

On the Generalization of SFT: A Reinforcement Learning Perspective with Reward Rectification

Flatter Tokens are More Valuable for Speculative Draft Model Training

UNITE: Universal kNowledge Integration from Task-specific Experts

Towards Understanding Feature Learning in Parameter Transfer

ALT: Adaptive Low‑Rank Transformation for Learngene-based Transformer Initialization

FLASH: Latent-Aware Semi-Autoregressive Speculative Decoding for Multimodal Tasks

d$^2$Cache: Accelerating Diffusion-Based LLMs via Dual Adaptive Caching

GeneLLM: Inheriting 1.25% of MoE-LLMs to Build Models of 8% Size that Retain 80% Performance

Constructing Variable-depth ViTs from Repeatable Low-bit Learngene

Chain-of-Learngene: A Scalable Learngene-based Paradigm for Building and Initializing Variable-Sized Language Models

LayerCake: Token-Aware Contrastive Decoding within Large Language Model Layers

Inheriting Generalizable Knowledge from LLMs to Diverse Vertical Tasks

Rethinking LLM Ensembling from the Perspective of Mixture Models

Aligning Draft and Target in Speculative Decoding: A CoT-Aware and SR-Guided Mixed Framework

Adversarial Graph Fusion for Incomplete Multi-view Semi-supervised Learning with Tensorial Imputation

FlowPrune: Accelerating Attention Flow Calculation by Pruning Flow Network

Fast Large Language Model Collaborative Decoding via Speculation

GraphIC: A Graph-Based In-Context Example Retrieval Model for Multi-Step Reasoning

Learngene Tells You How to Customize: Task-Aware Parameter Initialization at Flexible Scales

Learngene Tells You How to Customize: Task-Aware Parameter Prediction at Flexible Scales