Ax Yujia Zheng, Fan Feng, Yuke Li, Shaoan Xie, Kevin Murphy, Kun Zhang 5/14/2026

From Generalist to Specialist Representation

Nonparametric framework for learning task-relevant specialist representations from generalist models with identifiability analysis.

Ax Joana Pasquali, Ramiro N. Barros, Arthur S. Bianchessi, Vin\'icius Conte Turani, Jo\~ao Vitor Boer Abitante, Rafaela Cappelari Ravazio, Christian Mattjie, Ot\'avio Parraga, Lucas S. Kupssinsk\"u, Rodrigo C. Barros 5/14/2026

Low-Rank Adapters Initialization via Gradient Surgery for Continual Learning

LoRA initialization method using gradient surgery to mitigate catastrophic forgetting in continual LLM fine-tuning.

Ax Vin\'icius Conte Turani, Ot\'avio Parraga, Jo\~ao Vitor Boer Abitante, Kristen K. Arguello, Joana Pasquali, Ramiro N. Barros, Flavio du Pin Calmon, Christian Mattjie, Rodrigo C. Barros, Lucas S. Kupssinsk\"u 5/14/2026

Inference-Time Machine Unlearning via Gated Activation Redirection

Inference-time machine unlearning via gated activation redirection to remove memorized training data from LLMs without retraining.

Ax Yunshu Wu, Jiayi Cheng, Longxuan Yu, Partha Thakuria, Rob Brekelmans, Evangelos E. Papalexakis, Greg Ver Steeg 5/14/2026

Discrete Stochastic Localization for Non-autoregressive Generation

Discrete Stochastic Localization framework for non-autoregressive sequence generation using continuous diffusion with unit-sphere embeddings.

Ax Kaiyang Li, Shaobo Han, Qing Su, Shihao Ji 5/14/2026

Bayesian Model Merging

Bayesian approach to merging multiple task-specific expert models without retraining, using strong anchor models as inductive bias.

Ax Siyuan Liu (IIIS, Tsinghua University), Tinghong Chen (College of AI, Tsinghua University,Shanghai Qi Zhi Institute), Xinghan Li (IIIS, Tsinghua University), Yifei Wang (Amazon AGI SF Lab), Jingzhao Zhang (IIIS, Tsinghua University,Shanghai Qi Zhi Institute) 5/14/2026

Data Difficulty and the Generalization--Extrapolation Tradeoff in LLM Fine-Tuning

Systematic empirical and theoretical analysis of how data difficulty affects generalization and extrapolation in LLM fine-tuning.

Ax Changhao Li, Rushi Qiang, Jiawei Huang, Chenxiao Gao, Chao Zhang, Niao He, Bo Dai 5/14/2026

Revisiting DAgger in the Era of LLM-Agents

Study of DAgger algorithm for training long-horizon LLM agents, addressing covariate shift in multi-turn interactions with teacher supervision.

Ax Celine Lee, Jing Nathan Yan, Chen Liang, Jiaxin Shi, Yin Zhang, Jeremiah Liu, Pengcheng Yin, Fernando Pereira, Ed Chi, Derek Cheng, Alexander M. Rush, Ruoxi Wang 5/14/2026

The Efficiency Gap in Byte Modeling

Study of efficiency gap in byte-level language modeling comparing byte-level and masked diffusion approaches to traditional subword tokenization.

Ax Yanggan Gu, Shuo Cai, Zihao Wang, Wenjun Wang, Yuanyi Wang, Pengkai Wang, Sirui Huang, Su Lu, Jianmin Wu, Hongxia Yang 5/14/2026

FeatCal: Feature Calibration for Post-Merging Models

FeatCal addresses performance gaps in merged models by analyzing feature drift between merged and expert models, proposing methods to calibrate features during model merging.

Ax Zihan Guan, Qiao Jin, Guangzhi Xiong, Fangyuan Chen, Mengxuan Hu, Qingyu Chen, Yifan Peng, Zhiyong Lu, Anil Vullikanti 5/14/2026

Large Language Models Lack Temporal Awareness of Medical Knowledge

Evaluation showing LLMs lack temporal awareness in medical knowledge because benchmarks are atemporal while medical knowledge continuously evolves.