Ang Li

Capacity-Aware Inference: Mitigating the Straggler Effect in Mixture of Experts

CogniPair: GNWT-inspired cognitive architecture for generative agents for Social Pairing - Dating & Hiring Applications

What Is Preference Optimization Doing, How and Why?

Learning Flexible Large Multimodal Models with Arbitrary Modality Combinations

MindCraft: How Concept Trees Take Shape In Deep Models

Predict Overcharging: Auditing LLM APIs via Reasoning Length Estimation

VeriReason: Reinforcement Learning with Testbench Feedback for Reasoning-Enhanced Verilog Generation

Dense Video Understanding with Gated Residual Tokenization

Structure before the Machine: Input Space is the Prerequisite for Concepts

CoIn: Counting the Invisible Reasoning Tokens in Commercial Opaque LLM APIs

XYZ-Text2SQL-R1: Simple Rewards, Strong Reasoning in Text-to-SQL

Slim and Sparse: Towards Efficient Understanding–Generation Models

SymRTLO: Enhancing RTL Code Optimization with LLMs and Neuron-Inspired Symbolic Reasoning

Speculate, then Collaborate: Fusing Knowledge of Language Models during Decoding

Towards counterfactual fairness through auxiliary variables

What Matters in Transformers? Not All Attention is Needed

Speculate, then Collaborate: Fusing Knowledge of Language Models during Decoding

One Communication Round is All It Needs for Federated Fine-Tuning Foundation Models

Towards Efficient Mixture of Experts: A Holistic Study of Compression Techniques

ShareLoRA: Less Tuning, More Performance for LoRA Fine-tuning of LLMs

Safeguarding System Prompts: A Surrogate-Based Defense Against Injection Attacks