Hao Cui

Ke Xu, Yixin Wang, Zhongcheng Li et al.

Elastic precision quantization enables multi-bit deployment via a single optimization pass, fitting diverse quantization scenarios.Yet, the high storage and optimization costs associated with the Transformer architecture, research on elastic quantization remains limited, particularly for large language models.This paper proposes QuEPT, an efficient post-training scheme that reconstructs block-wise multi-bit errors with one-shot calibration on a small data slice. It can dynamically adapt to various predefined bit-widths by cascading different low-rank adapters, and supports real-time switching between uniform quantization and mixed precision quantization without repeated optimization. To enhance accuracy and robustness, we introduce Multi-Bit Token Merging (MB-ToMe) to dynamically fuse token features across different bit-widths, improving robustness during bit-width switching. Additionally, we propose Multi-Bit Cascaded Low-Rank adapters (MB-CLoRA) to strengthen correlations between bit-width groups, further improve the overall performance of QuEPT. Extensive experiments demonstrate that QuEPT achieves comparable or better performance to existing state-of-the-art post-training quantization methods.Our code is available at https://github.com/xuke225/QuEPT

Hao Cui, Taha Yasseri

As AI becomes more embedded in workplaces, it is shifting from a tool for efficiency to an active force in organizational decision-making. Whether due to anthropomorphism or intentional design choices, people often assign human-like qualities, including gender, to AI systems. However, how AI managers are perceived in comparison to human managers and how gender influences these perceptions remains uncertain. To investigate this, we conducted randomized controlled trials (RCTs) where teams of three participants worked together under a randomly assigned manager. The manager was either a human or an AI and was presented as male, female, or gender-unspecified. The manager's role was to select the best-performing team member for an additional award. Our findings reveal that while participants initially showed no strong preference based on manager type or gender, their perceptions changed notably after experiencing the award process. As expected, those who received awards rated their managers as more trustworthy, competent, and fair, and they were more willing to work with similar managers in the future. In contrast, those who were not selected viewed them less favorably. However, male managers, whether human or AI, were more positively received by awarded participants, whereas female managers, especially female AI managers, faced greater skepticism and negative judgments when they did not give awards. These results suggest that gender bias in leadership extends beyond human managers to include AI-driven decision-makers as well. As AI assumes more managerial responsibilities, understanding and addressing these biases will be crucial for designing fair and effective AI management systems.

Hao Cui, Zahra Shamsi, Gowoon Cheon et al.

Scientific problem-solving involves synthesizing information while applying expert knowledge. We introduce CURIE, a scientific long-Context Understanding,Reasoning and Information Extraction benchmark to measure the potential of Large Language Models (LLMs) in scientific problem-solving and assisting scientists in realistic workflows. This benchmark introduces ten challenging tasks with a total of 580 problems and solution pairs curated by experts in six disciplines - materials science, condensed matter physics, quantum computing, geospatial analysis, biodiversity, and proteins - covering both experimental and theoretical work-flows in science. We evaluate a range of closed and open LLMs on tasks in CURIE which requires domain expertise, comprehension of long in-context information,and multi-step reasoning. While Gemini Flash 2.0 and Claude-3 show consistent high comprehension across domains, the popular GPT-4o and command-R+ fail dramatically on protein sequencing tasks. With the best performance at 32% there is much room for improvement for all models. We hope that insights gained from CURIE can guide the future development of LLMs in sciences. Evaluation code and data are in https://github.com/google/curie

Nayantara Mudur, Hao Cui, Subhashini Venugopalan et al.

Building precise simulations of the real world and invoking numerical solvers to answer quantitative problems is an essential requirement in engineering and science. We present FEABench, a benchmark to evaluate the ability of large language models (LLMs) and LLM agents to simulate and solve physics, mathematics and engineering problems using finite element analysis (FEA). We introduce a comprehensive evaluation scheme to investigate the ability of LLMs to solve these problems end-to-end by reasoning over natural language problem descriptions and operating COMSOL Multiphysics$^\circledR$, an FEA software, to compute the answers. We additionally design a language model agent equipped with the ability to interact with the software through its Application Programming Interface (API), examine its outputs and use tools to improve its solutions over multiple iterations. Our best performing strategy generates executable API calls 88% of the time. LLMs that can successfully interact with and operate FEA software to solve problems such as those in our benchmark would push the frontiers of automation in engineering. Acquiring this capability would augment LLMs' reasoning skills with the precision of numerical solvers and advance the development of autonomous systems that can tackle complex problems in the real world. The code is available at https://github.com/google/feabench

Di Zhang, Zhangpeng Gong, Xiaobo Pang et al.

Foundation models have recently achieved impressive success in computational pathology, demonstrating strong generalization across diverse histopathology tasks. However, existing models overlook the heterogeneous and non-uniform organization of pathological regions of interest (ROIs) because they rely on natural image backbones not tailored for tissue morphology. Consequently, they often fail to capture the coherent tissue architecture beyond isolated patches, limiting interpretability and clinical relevance. To address these challenges, we present Cross-modal Adaptive Region Encoder (CARE), a foundation model for pathology that automatically partitions WSIs into several morphologically relevant regions. Specifically, CARE employs a two-stage pretraining strategy: (1) a self-supervised unimodal pretraining stage that learns morphological representations from 34,277 whole-slide images (WSIs) without segmentation annotations, and (2) a cross-modal alignment stage that leverages RNA and protein profiles to refine the construction and representation of adaptive regions. This molecular guidance enables CARE to identify biologically relevant patterns and generate irregular yet coherent tissue regions, selecting the most representative area as ROI. CARE supports a broad range of pathology-related tasks, using either the ROI feature or the slide-level feature obtained by aggregating adaptive regions. Based on only one-tenth of the pretraining data typically used by mainstream foundation models, CARE achieves superior average performance across 33 downstream benchmarks, including morphological classification, molecular prediction, and survival analysis, and outperforms other foundation model baselines overall.

Hao Cui, Taha Yasseri

Current societal challenges exceed the capacity of humans operating either alone or collectively. As AI evolves, its role within human collectives will vary from an assistive tool to a participatory member. Humans and AI possess complementary capabilities that, together, can surpass the collective intelligence of either humans or AI in isolation. However, the interactions in human-AI systems are inherently complex, involving intricate processes and interdependencies. This review incorporates perspectives from complex network science to conceptualize a multilayer representation of human-AI collective intelligence, comprising cognition, physical, and information layers. Within this multilayer network, humans and AI agents exhibit varying characteristics; humans differ in diversity from surface-level to deep-level attributes, while AI agents range in degrees of functionality and anthropomorphism. We explore how agents' diversity and interactions influence the system's collective intelligence and analyze real-world instances of AI-enhanced collective intelligence. We conclude by considering potential challenges and future developments in this field.

Eser Aygün, Anastasiya Belyaeva, Gheorghe Comanici et al.

The cycle of scientific discovery is frequently bottlenecked by the slow, manual creation of software to support computational experiments. To address this, we present an AI system that creates expert-level scientific software whose goal is to maximize a quality metric. The system uses a Large Language Model (LLM) and Tree Search (TS) to systematically improve the quality metric and intelligently navigate the large space of possible solutions. The system achieves expert-level results when it explores and integrates complex research ideas from external sources. The effectiveness of tree search is demonstrated across a wide range of benchmarks. In bioinformatics, it discovered 40 novel methods for single-cell data analysis that outperformed the top human-developed methods on a public leaderboard. In epidemiology, it generated 14 models that outperformed the CDC ensemble and all other individual models for forecasting COVID-19 hospitalizations. Our method also produced state-of-the-art software for geospatial analysis, neural activity prediction in zebrafish, time series forecasting and numerical solution of integrals. By devising and implementing novel solutions to diverse tasks, the system represents a significant step towards accelerating scientific progress.

Rahmadi Trimananda, Hieu Le, Hao Cui et al.

Virtual reality (VR) is an emerging technology that enables new applications but also introduces privacy risks. In this paper, we focus on Oculus VR (OVR), the leading platform in the VR space and we provide the first comprehensive analysis of personal data exposed by OVR apps and the platform itself, from a combined networking and privacy policy perspective. We experimented with the Quest 2 headset and tested the most popular VR apps available on the official Oculus and the SideQuest app stores. We developed OVRseen, a methodology and system for collecting, analyzing, and comparing network traffic and privacy policies on OVR. On the networking side, we captured and decrypted network traffic of VR apps, which was previously not possible on OVR, and we extracted data flows, defined as <app, data type, destination>. Compared to the mobile and other app ecosystems, we found OVR to be more centralized and driven by tracking and analytics, rather than by third-party advertising. We show that the data types exposed by VR apps include personally identifiable information (PII), device information that can be used for fingerprinting, and VR-specific data types. By comparing the data flows found in the network traffic with statements made in the apps' privacy policies, we found that approximately 70% of OVR data flows were not properly disclosed. Furthermore, we extracted additional context from the privacy policies, and we observed that 69% of the data flows were used for purposes unrelated to the core functionality of apps.

Lantian Li, Ruiqi Liu, Jiawen Kang et al.

Research on speaker recognition is extending to address the vulnerability in the wild conditions, among which genre mismatch is perhaps the most challenging, for instance, enrollment with reading speech while testing with conversational or singing audio. This mismatch leads to complex and composite inter-session variations, both intrinsic (i.e., speaking style, physiological status) and extrinsic (i.e., recording device, background noise). Unfortunately, the few existing multi-genre corpora are not only limited in size but are also recorded under controlled conditions, which cannot support conclusive research on the multi-genre problem. In this work, we firstly publish CN-Celeb, a large-scale multi-genre corpus that includes in-the-wild speech utterances of 3,000 speakers in 11 different genres. Secondly, using this dataset, we conduct a comprehensive study on the multi-genre phenomenon, in particular the impact of the multi-genre challenge on speaker recognition and the performance gain when the new dataset is used to conduct multi-genre training.

Jie Zhang, Dongdong Chen, Jing Liao et al.

Deep learning has achieved tremendous success in numerous industrial applications. As training a good model often needs massive high-quality data and computation resources, the learned models often have significant business values. However, these valuable deep models are exposed to a huge risk of infringements. For example, if the attacker has the full information of one target model including the network structure and weights, the model can be easily finetuned on new datasets. Even if the attacker can only access the output of the target model, he/she can still train another similar surrogate model by generating a large scale of input-output training pairs. How to protect the intellectual property of deep models is a very important but seriously under-researched problem. There are a few recent attempts at classification network protection only. In this paper, we propose the first model watermarking framework for protecting image processing models. To achieve this goal, we leverage the spatial invisible watermarking mechanism. Specifically, given a black-box target model, a unified and invisible watermark is hidden into its outputs, which can be regarded as a special task-agnostic barrier. In this way, when the attacker trains one surrogate model by using the input-output pairs of the target model, the hidden watermark will be learned and extracted afterward. To enable watermarks from binary bits to high-resolution images, both traditional and deep spatial invisible watermarking mechanism are considered. Experiments demonstrate the robustness of the proposed watermarking mechanism, which can resist surrogate models learned with different network structures and objective functions. Besides deep models, the proposed method is also easy to be extended to protect data and traditional image processing algorithms.