Kai Yu

Full Professor@Shanghai Jiao Tong University·中国·OpenReview

研究方向

deep learning · question and answering · natural language processing · dialogue systems · reinforcement learning · spoken language understanding · speech synthesis · speech recognition · language modelling · speaker recognition

Kai Yu

Diffusion LLMs Can Do Faster-Than-AR Inference via Discrete Diffusion Forcing

DiSRouter: Distributed Self-Routing for LLM Selections

Decomposing Scientific Paper Queries with Draft-and-Follow Policy Optimization to Narrow Knowing-Doing Gap

Omni-Captioner: Data Pipeline, Models, and Benchmark for Omni Detailed Perception

AirQA: A Comprehensive QA Dataset for AI Research with Instance-Level Evaluation

Empowering LLM Tool Invocation with Tool-call Reward Model

Done Is Better than Perfect: Unlocking Efficient Reasoning by Structured Multi-Turn Decomposition

ChemDFM-R: A Chemical Reasoning LLM Enhanced with Atomized Chemical Knowledge

UltraVoice: Scaling Fine-Grained Style-Controlled Speech Conversations for Spoken Dialogue Models

MMWebGen: Benchmarking Multimodal Webpage Generation

MS-BART: Unified Modeling of Mass Spectra and Molecules for Structure Elucidation

Task-Specific Data Selection for Instruction Tuning via Monosemantic Neuronal Activations

Enhancing LLM Reliability via Explicit Knowledge Boundary Modeling

Communication-Efficient Diffusion Denoising Parallelization via Reuse-then-Predict Mechanism

Heads up! Large Language Models Can Perform Tasks Without Your Instruction via Selective Attention Head Masking

F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching

Reducing Tool Hallucination via Reliability Alignment

Boosting Long-Context LLM Inference Efficiency with Intra-Layer Attention Similarity

GigaSpeech 2: An Evolving, Large-Scale and Multi-domain ASR Corpus for Low-Resource Languages with Automated Crawling, Transcription and Refinement