Zhouchen Lin

Online Pseudo-Zeroth-Order Training of Neuromorphic Spiking Neural Networks

DistDF: Time-series Forecasting Needs Joint-distribution Wasserstein Alignment

Quadratic Direct Forecast for Training Multi-Step Time-Series Forecast Models

LogiConBench: Benchmarking Logical Consistencies of LLMs

DNT: a Deeply Normalized Transformer that can be trained by Momentum SGD

ProofRM: A Scalable Pipeline to Train a Generalized Math Proof Reward Model

Rethinking the Flow-based Gradual Domain Adaption: A Semi-Dual Transport Perspective

Transformers with Endogenous In-Context Learning: Bias Characterization and Mitigation

Partial Identification via Optimal Transport under Complex Constraints on Treatments and Potential Outcome Measures

VACT: A Video Automatic Causal Testing System and a Benchmark

Enhancing Logical Reasoning of Large Language Models via Phased Fine-Tuning

Governing Equation Discovery from Data Based on Differential Invariants

Personal Tokens Matter: Towards Token-Aware Training for Personalized LLMs

Conda: Column-Normalized Adam for Training Large Language Models Faster

GL-Fusion: Rethinking the Combination of Graph Neural Network and Large Language model

Simple Convergence Proof of Adam From a Sign-like Descent Perspective

Projective Equivariant Networks via Second-order Fundamental Differential Invariants

Time-o1: Time-Series Forecasting Needs Transformed Label Alignment

On the $O(\frac{\sqrt{d}}{K^{1/4}})$ Convergence Rate of AdamW Measured by $\ell_1$ Norm

Stepsize anything: A unified learning rate schedule for budgeted-iteration training

AdaMSS: Adaptive Multi-Subspace Approach for Parameter-Efficient Fine-Tuning

Active Treatment Effect Estimation via Limited Samples

Pyramidal Flow Matching for Efficient Video Generative Modeling

PaZO: Preconditioned Accelerated Zeroth-Order Optimization for Fine-Tuning LLMs

Improving Model Representation and Reducing KV Cache via Skip Connections with First Value Heads

Inverse Methods for Missing Data Imputation

Unbiased Recommender Learning from Implicit Feedback via Weakly Supervised Learning

Number Cookbook: Number Understanding of Language Models and How to Improve It

TC-MoE: Augmenting Mixture of Experts with Ternary Expert Choice

Language Ranker: A Lightweight Ranking framework for LLM Decoding

Explicit Discovery of Nonlinear Symmetries from Dynamic Data

SEPARATE: A Simple Low-rank Projection for Gradient Compression in Modern Large-scale Model Training Process

Tool Decoding: A Plug-and-Play Approach to Enhancing Language Models for Tool Usage

PseuZO: Pseudo-Zeroth-Order Algorithm for Training Deep Neural Networks

Affine Steerable Equivariant Layer for Canonicalization of Neural Networks

Low-Dimension-to-High-Dimension Generalization and Its Implications for Length Generalization

MLAE: Masked LoRA Experts for Visual Parameter-Efficient Fine-Tuning

GL-Fusion: Rethinking the Combination of Graph Neural Network and Large Language model

Low-Dimension-to-High-Dimension Generalization and Its Implications for Length Generalization

Finding Second-order Stationary Points for Generalized-Smooth Nonconvex Minimax Optimization via Gradient-based Algorithm

Provable Faster Zeroth-order Method for Bilevel Optimization with Optimal Dependency on Error and Dimension

Incorporating Arbitrary Matrix Group Equivariance into KANs

EKAN: Equivariant Kolmogorov-Arnold Networks

Variance-Reduced Normalized Zeroth Order Method for Generalized-Smooth Non-Convex Optimization