Publications (18)
Generalizing from SIMPLE to HARD Visual Reasoning: Can We Mitigate Modality Imbalance in VLMs?
Simon Park, Abhishek Panigrahi, Yun Cheng +3
Vision Language Models (VLMs) are impressive at visual question answering and image captioning. But they underperform on multi-step visual reasoning -- even compared to LLMs on the…
Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates
Kaifeng Lyu, Haoyu Zhao, Xinran Gu +3
Public LLMs such as the Llama 2-Chat underwent alignment training and were considered safe. Recently Qi et al. [2024] reported that even benign fine-tuning on seemingly safe datase…
Characterization of Group-Strategyproof Mechanisms for Facility Location in Strictly Convex Space
Pingzhong Tang, Dingli Yu, Shengyu Zhao
We characterize the class of group-strategyproof mechanisms for the single facility location game in any unconstrained strictly convex space. A mechanism is \emph{group-strategypro…
A partisan districting protocol with provably nonpartisan outcomes
Wesley Pegden, Ariel D. Procaccia, Dingli Yu
We design and analyze a protocol for dividing a state into districts, where parties take turns proposing a division, and freezing a district from the other party's proposed divisio…
New Definitions and Evaluations for Saliency Methods: Staying Intrinsic, Complete and Sound
Arushi Gupta, Nikunj Saunshi, Dingli Yu +2
Saliency methods compute heat maps that highlight portions of an input that were most {\em important} for the label assigned to it by a deep net. Evaluations of saliency methods co…
Phi-4 Technical Report
Marah Abdin, Jyoti Aneja, Harkirat Behl +24
We present phi-4, a 14-billion parameter language model developed with a training recipe that is centrally focused on data quality. Unlike most language models, where pre-training…
Harnessing the Power of Infinitely Wide Deep Nets on Small-data Tasks
Sanjeev Arora, Simon S. Du, Zhiyuan Li +3
Recent research shows that the following two models are equivalent: (a) infinitely wide neural networks (NNs) trained under l2 loss by gradient descent with infinitesimally small l…
Tensor Programs VI: Feature Learning in Infinite-Depth Neural Networks
Greg Yang, Dingli Yu, Chen Zhu +1
By classifying infinite-width neural networks and identifying the *optimal* limit, Tensor Programs IV and V demonstrated a universal way, called P, for *widthwise hyperparamete…
A Geometric Analysis-Based Safety Assessment Framework for Marine Vehicle Route Decision-Making
Zilong Xu, Zihao Wang, He Li +3
This paper develops a Geometric Analysis-based Route Safety Assessment (GARSA) framework to enhance the safety of marine vehicles navigating in restricted waters. Utilizing line an…
Enhanced Convolutional Neural Tangent Kernels
Zhiyuan Li, Ruosong Wang, Dingli Yu +4
Recent research shows that for training with loss, convolutional neural networks (CNNs) whose width (number of channels in convolutional layers) goes to infinity correspon…
OpenAI GPT-5 System Card
Aaditya Singh, Adam Fry, Adam Perelman +483
This is the system card published alongside the OpenAI GPT-5 launch, August 2025. GPT-5 is a unified system with a smart and fast model that answers most questions, a deeper reason…
AI-Assisted Generation of Difficult Math Questions
Vedant Shah, Dingli Yu, Kaifeng Lyu +8
Current LLM training positions mathematical reasoning as a core capability. With publicly available sources fully tapped, there is unmet demand for diverse and challenging math que…
Weak-to-Strong Generalization Even in Random Feature Networks, Provably
Marko Medvedev, Kaifeng Lyu, Dingli Yu +3
Weak-to-Strong Generalization (Burns et al., 2024) is the phenomenon whereby a strong student, say GPT-4, learns a task from a weak teacher, say GPT-2, and ends up significantly ou…
Skill-Mix: a Flexible and Expandable Family of Evaluations for AI models
Dingli Yu, Simran Kaur, Arushi Gupta +3
With LLMs shifting their role from statistical modeling of language to serving as general-purpose AI agents, how should LLM evaluations change? Arguably, a key ability of an AI age…
Simple and Effective Regularization Methods for Training on Noisily Labeled Data with Generalization Guarantee
Wei Hu, Zhiyuan Li, Dingli Yu
Over-parameterized deep neural networks trained by simple first-order methods are known to be able to fit any labeling of data. Such over-fitting ability hinders generalization whe…
A Kernel-Based View of Language Model Fine-Tuning
Sadhika Malladi, Alexander Wettig, Dingli Yu +2
It has become standard to solve NLP tasks by fine-tuning pre-trained language models (LMs), especially in low-data settings. There is minimal theoretical understanding of empirical…
ConceptMix: A Compositional Image Generation Benchmark with Controllable Difficulty
Xindi Wu, Dingli Yu, Yangsibo Huang +2
Compositionality is a critical capability in Text-to-Image (T2I) models, as it reflects their ability to understand and combine multiple concepts from text descriptions. Existing e…
Can Models Learn Skill Composition from Examples?
Haoyu Zhao, Simran Kaur, Dingli Yu +2
As large language models (LLMs) become increasingly advanced, their ability to exhibit compositional generalization -- the capacity to combine learned skills in novel ways not enco…