Ax Jaehun Jung, Hyunwoo Kim, Brandon Cui, Ximing Lu, David Acuna, Prithviraj Ammanabrolu, Yejin Choi 5/18/2026

DeltaPrompts: Escaping the Zero-Delta Trap in Multimodal Distillation

DeltaPrompts improves distillation of Vision-Language Models by identifying ineffective prompts in training data. Addresses zero-delta prompt problem in multimodal model compression.

Ax Haizhong Zheng, Yizhuo Di, Jiahui Wang, Shuowei Jin, Xueshen Liu, Yongji Wu, Z. Morley Mao, Ion Stoica, Jiawei Zhao, Beidi Chen 5/18/2026

AstraFlow: Dataflow-Oriented Reinforcement Learning for Agentic LLMs

AstraFlow is dataflow-oriented system for scaling reinforcement learning on agentic LLMs across multi-policy training on heterogeneous compute.

Ax Zezhong Ding, Jin Li, Xugang Wang, Xike Xie 5/18/2026

Gaussian Relational Graph Transformer

GelGT combines Gaussian processes with graph transformers to jointly model structural, semantic, and temporal information in relational data.

Ax Aditya Kudre, Heng-Sheng Chang, Prashant G. Mehta 5/18/2026

Transformer-like Inference from Optimal Control

Derives transformer-like inference architectures from optimal control theory, recovering decoder-only transformer operations from first principles.

Ax Vishy Gopal, Aris Ilias Goutis, Ralph Crewe, Erin Yanacek, Rorry Brenner 5/18/2026

Perforated Neural Networks for Keyword Spotting

Perforated backpropagation applied to keyword spotting enables simultaneous improvements in edge model accuracy and size under strict memory constraints.

Ax Jaeseung Heo, Kyeongheung Yun, Youngbin Choi, Sehyun Hwang, Jungseul Ok, Dongwoo Kim 5/18/2026

Interaction-Aware Influence Functions for Group Attribution

Interaction-aware influence functions extend training example attribution to groups, capturing redundancy and complementarity beyond summed individual influences.

Ax Yuan Zhang, Lifeng Guo, Junwen Pan, Chang Liu, Wenzhao Zheng, Kuan Cheng, Kurt Keutzer, Shanghang Zhang 5/18/2026

SEED: Targeted Data Selection by Weighted Independent Set

SEED formulates data selection as weighted independent set problem to identify compact, high-quality, diverse subsets from training corpora.

Ax An Nguyen, Jaesik Choi, Anh Tong 5/18/2026

LoCO: Low-rank Compositional Rotation Fine-tuning

LoCO: parameter-efficient fine-tuning method for foundation models using low-rank compositional orthogonal updates while preserving geometric structure.