Researcher — paper.dinq.me

2026 顶会论文

作者位次 #3/3ICLR 2026Poster· Optimization / Non-Convex

A Convergence Analysis of Adaptive Optimizers under Floating-point Quantization

Xuan Tang · Jichu Li · Difan Zou

效率与系统OpenReview

作者位次 #2/3ICLR 2026Poster· Deep Learning / Attention Mechanisms

Hyper-SET: Designing Transformers via Hyperspherical Energy Minimization

Yunzhe Hu · Difan Zou · Dong Xu

可解释性OpenReview

作者位次 #4/4ICLR 2026Poster· Deep Learning / Everything Else

Does Higher Interpretability Imply Better Utility? A Pairwise Analysis on Sparse Autoencoders

Xu Wang · Yan Hu · Wang Benyou · Difan Zou

可解释性OpenReview

作者位次 #3/3ICLR 2026Poster· Theory / Learning Theory

Learning under Quantization for High-Dimensional Linear Regression

Dechen Zhang · Junwei Su · Difan Zou

效率与系统理论与学习动力学OpenReview

作者位次 #7/11ICLR 2026Poster· Deep Learning / Generative Models and Autoencoders

Routing Matters in MoE: Scaling Diffusion Transformers with Explicit Routing Guidance

Yujie Wei · Shiwei Zhang · Hangjie Yuan · Yujin Han · Zhekai Chen · Jiayu Wang · Difan Zou · Xihui Liu · yingya zhang · Yu Liu · Hongming Shan

效率与系统OpenReview

作者位次 #3/3ICLR 2026Poster· Theory / Interpretability and Visualization

Reshaping Reasoning in LLMs: A Theoretical Analysis of RL Training Dynamics through Pattern Selection

Xingwu Chen · Tianle Li · Difan Zou

推理可解释性多模态OpenReview

作者位次 #8/8ICLR 2026Poster

Turning Internal Gap into Self-Improvement: Promoting the Generation-Understanding Unification in MLLMs

Yujin Han · Hao Chen · Andi Han · Zhiheng Wang · Xinyu Liu · yingya zhang · Shiwei Zhang · Difan Zou

强化学习与后训练OpenReview

作者位次 #4/4ACL 2026Findings· Findings of ACL

Breaking Contextual Inertia: Reinforcement Learning with Single-Turn Anchors for Stable Multi-Turn Interaction

Xingwu Chen · Zhanqiu Zhang · Steven Y. Guo · Difan Zou

摘要

While LLMs demonstrate strong reasoning capabilities when provided with full information in a single turn, they exhibit substantial vulnerability in multi-turn interactions. Specifically, when information is revealed incrementally or requires updates, models frequently fail to integrate new const...

推理强化学习与后训练对齐与偏好OpenReview 项目页

作者位次 #3/4AAAI 2026Poster· Artificial Intelligence

SIDE: Surrogate Conditional Data Extraction from Diffusion Models

Yunhao Chen · Shujie Wang · Difan Zou · Xingjun Ma

OpenReview

作者位次 #3/4AAAI 2026Poster· Artificial Intelligence

Learning Diffusion Policy from Primitive Skills for Robot Manipulation

Zhihao Gu · Ming Yang · Difan Zou · Dong Xu