Ax Shengkun Tang, Zekun Wang, Bo Zheng, Liangyu Wang, Rui Men, Siqi Zhang, Xiulong Yuan, Zihan Qiu, Zhiqiang Shen, Dayiheng Liu 5/12/2026

SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training

Systematically studies compression techniques (pruning and knowledge distillation) for large-scale MoE model pretraining, examining initialization quality and expert compression strategies.

Ax Jiawei Lian, Jianhong Pan, Lefan Wang, Yi Wang, Tairan Huang, Shaohui Mei, Lap-Pui Chau 5/12/2026

LLM-Agnostic Semantic Representation Attack

LLM-Agnostic semantic representation attack circumvents alignment safeguards by optimizing adversarial prompts at semantic level rather than token level.

Ax Zongmin Yu, Liu Yang 5/12/2026

Evolutionary Ensemble of Agents

Evolutionary Ensemble organizes coding agents into decentralized co-evolving system for algorithmic discovery by evolving cumulative guidance and learned skills.

Ax Haejoon Lee, Vincent-Daniel Yun, Hyeonho Oh, Dimitra Panagou, Sai Praneeth Karimireddy 5/12/2026

Robust Multi-Agent LLMs under Byzantine Faults

Research on defending multi-agent LLM systems against Byzantine faults in peer-to-peer networks to prevent adversarial manipulation and system degradation.

Ax Karim Othman, Jonas Petersen, Matei Ignuta-Ciuncanu, Riccardo Maggioni, Camilla Mazzoleni, Federico Martelli, Philipp Petersen 5/12/2026

FactoryNet: A Large-Scale Dataset toward Industrial Time-Series Foundation Models

FactoryNet introduces a 51M datapoint dataset for industrial time-series foundation models with a unified S-E-F-C schema for zero-shot cross-embodiment transfer and anomaly detection.

Ax Zhiyang Dou, Minghao Guo, Haixu Wu, Doug Roble, Tuur Stuyck, Wojciech Matusik 5/12/2026

RigidFormer: Learning Rigid Dynamics using Transformers

RigidFormer uses transformers to learn rigid-body dynamics from mesh-free representations like point clouds, improving computational efficiency over vertex-level methods.

Ax Sepehr Harfi, Ahmad Salimi, Dongming Shen, Alex Smola 5/12/2026

ProactBench: Beyond What The User Asked For

ProactBench evaluates LLM conversational proactivity: ability to infer and act on implied user needs beyond explicit requests.

Ax Juanwu Lu, Ziran Wang 5/12/2026

On Variance Reduction in Learning Mean Flows

Theoretical analysis of variance reduction in MeanFlow one-step generative modeling, addressing training instability and gradient variance.

Ax Jeongho Bang, Marcin Paw{\l}owski 5/12/2026

Neural Information Causality

Theoretical framework embedding information causality into representation learning via query-separated computation, tangential to core interests.