Wanli Ouyang
YOU?
Author Swipe
View article: Evidential deep learning for interatomic potentials
Evidential deep learning for interatomic potentials Open
Machine learning interatomic potentials have been widely used to facilitate large-scale molecular simulations with accuracy comparable to ab initio methods. To ensure the reliability of the simulation, the training dataset is iteratively e…
Glasses-free 3D display with ultrawide viewing range using deep learning Open
Glasses-free three-dimensional (3D) displays provide users with an immersive visual experience without the need of any wearable devices1,2. To achieve high-quality 3D imaging, a display should have both large linear dimensions and a wide v…
View article: Climate science data can be compressed efficiently by dual-stage extreme compression with a variational auto-encoder transformer
Climate science data can be compressed efficiently by dual-stage extreme compression with a variational auto-encoder transformer Open
Climate change makes accurate weather prediction and large-scale data analysis increasingly crucial, but the sheer volume of weather data strains data storage and sharing. Here we introduce Aeolus, a deep learning framework with powerful V…
View article: CoMAS: Co-Evolving Multi-Agent Systems via Interaction Rewards
CoMAS: Co-Evolving Multi-Agent Systems via Interaction Rewards Open
Self-evolution is a central research topic in enabling large language model (LLM)-based agents to continually improve their capabilities after pretraining. Recent research has witnessed a transition from reinforcement learning (RL)-free to…
View article: Cache-to-Cache: Direct Semantic Communication Between Large Language Models
Cache-to-Cache: Direct Semantic Communication Between Large Language Models Open
Multi-LLM systems harness the complementary strengths of diverse Large Language Models, achieving performance and efficiency gains unattainable by a single model. In existing designs, LLMs communicate through text, forcing internal represe…
View article: A surrogate model based on parametric neural network solvers for laminar flows around aerofoils
A surrogate model based on parametric neural network solvers for laminar flows around aerofoils Open
Physics-informed neural networks (PINNs) have emerged as a popular approach for solving forward, inverse, and parametric problems involving partial differential equations. However, their performance is often limited by ill-conditioning in …
View article: Agentic Jigsaw Interaction Learning for Enhancing Visual Perception and Reasoning in Vision-Language Models
Agentic Jigsaw Interaction Learning for Enhancing Visual Perception and Reasoning in Vision-Language Models Open
Although current large Vision-Language Models (VLMs) have advanced in multimodal understanding and reasoning, their fundamental perceptual and reasoning abilities remain limited. Specifically, even on simple jigsaw tasks, existing VLMs per…
View article: PhysicsMinions: Winning Gold Medals in the Latest Physics Olympiads with a Coevolutionary Multimodal Multi-Agent System
PhysicsMinions: Winning Gold Medals in the Latest Physics Olympiads with a Coevolutionary Multimodal Multi-Agent System Open
Physics is central to understanding and shaping the real world, and the ability to solve physics problems is a key indicator of real-world physical intelligence. Physics Olympiads, renowned as the crown of competitive physics, provide a ri…
View article: Understand Before You Generate: Self-Guided Training for Autoregressive Image Generation
Understand Before You Generate: Self-Guided Training for Autoregressive Image Generation Open
Recent studies have demonstrated the importance of high-quality visual representations in image generation and have highlighted the limitations of generative models in image understanding. As a generative paradigm originally designed for n…
View article: ChemBOMAS: Accelerated BO in Chemistry with LLM-Enhanced Multi-Agent System
ChemBOMAS: Accelerated BO in Chemistry with LLM-Enhanced Multi-Agent System Open
Bayesian optimization (BO) is a powerful tool for scientific discovery in chemistry, yet its efficiency is often hampered by the sparse experimental data and vast search space. Here, we introduce ChemBOMAS: a large language model (LLM)-enh…
View article: Interleaving Reasoning for Better Text-to-Image Generation
Interleaving Reasoning for Better Text-to-Image Generation Open
Unified multimodal understanding and generation models recently have achieve significant improvement in image generation capability, yet a large gap remains in instruction following and detail preservation compared to systems that tightly …
View article: DispFormer: A Pretrained Transformer Incorporating Physical Constraints for Dispersion Curve Inversion
DispFormer: A Pretrained Transformer Incorporating Physical Constraints for Dispersion Curve Inversion Open
Surface wave dispersion curve inversion is crucial for estimating subsurface shear‐wave velocity , yet traditional methods often face challenges related to computational cost, non‐uniqueness, and sensitivity to initial models. While deep l…
View article: CMPhysBench: A Benchmark for Evaluating Large Language Models in Condensed Matter Physics
CMPhysBench: A Benchmark for Evaluating Large Language Models in Condensed Matter Physics Open
We introduce CMPhysBench, designed to assess the proficiency of Large Language Models (LLMs) in Condensed Matter Physics, as a novel Benchmark. CMPhysBench is composed of more than 520 graduate-level meticulously curated questions covering…
View article: Advancing Climate Science Data Efficiency by Dual-stage Extreme Compression with an Efficient VAEformer
Advancing Climate Science Data Efficiency by Dual-stage Extreme Compression with an Efficient VAEformer Open
The growing threat of climate change highlights the urgent need for deeper understanding and precise weather prediction. Despite advancements in atmospheric research, the ever-growing volume of weather data challenges efficient storage and…
View article: From AI for Science to Agentic Science: A Survey on Autonomous Scientific Discovery
From AI for Science to Agentic Science: A Survey on Autonomous Scientific Discovery Open
Artificial intelligence (AI) is reshaping scientific discovery, evolving from specialized computational tools into autonomous research partners. We position Agentic Science as a pivotal stage within the broader AI for Science paradigm, whe…
View article: Data-driven global ocean modeling for seasonal to decadal prediction
Data-driven global ocean modeling for seasonal to decadal prediction Open
Accurate modeling of ocean dynamics is crucial for enhancing our understanding of complex ocean circulation processes, predicting climate variability, and tackling challenges posed by climate change. Although great efforts have been made t…
View article: CTTS: Collective Test-Time Scaling
CTTS: Collective Test-Time Scaling Open
Test-time scaling (TTS) has emerged as a promising, training-free approach for enhancing large language model (LLM) performance. However, the efficacy of existing methods, such as Best-of-N and Self-Consistency, is fundamentally constraine…
View article: Fitness aligned structural modeling enables scalable virtual screening with AuroBind
Fitness aligned structural modeling enables scalable virtual screening with AuroBind Open
Most human proteins remain undrugged, over 96% of human proteins remain unexploited by approved therapeutics. While structure-based virtual screening promises to expand the druggable proteome, existing methods lack atomic-level precision a…
View article: A Self-Evolving AI Agent System for Climate Science
A Self-Evolving AI Agent System for Climate Science Open
Scientific progress in Earth science depends on integrating data across the planet's interconnected spheres. However, the accelerating volume and fragmentation of multi-sphere knowledge and data have surpassed human analytical capacity. Th…
View article: STAR: A Benchmark for Astronomical Star Fields Super-Resolution
STAR: A Benchmark for Astronomical Star Fields Super-Resolution Open
Super-resolution (SR) advances astronomical imaging by enabling cost-effective high-resolution capture, crucial for detecting faraway celestial objects and precise structural analysis. However, existing datasets for astronomical SR (ASR) e…
View article: Multimodal-Guided Dynamic Dataset Pruning for Robust and Efficient Data-Centric Learning
Multimodal-Guided Dynamic Dataset Pruning for Robust and Efficient Data-Centric Learning Open
Modern deep models are trained on large real-world datasets, where data quality varies and redundancy is common. Data-centric approaches such as dataset pruning have shown promise in improving training efficiency and model performance. How…
View article: Open-Source LLMs Collaboration Beats Closed-Source LLMs: A Scalable Multi-Agent System
Open-Source LLMs Collaboration Beats Closed-Source LLMs: A Scalable Multi-Agent System Open
This paper aims to demonstrate the potential and strengths of open-source collectives. It leads to a promising question: Can we harness multiple open-source LLMs to match or even beat the closed-source LLMs? To answer this, we propose SMAC…
View article: Reflections Unlock: Geometry-Aware Reflection Disentanglement in 3D Gaussian Splatting for Photorealistic Scenes Rendering
Reflections Unlock: Geometry-Aware Reflection Disentanglement in 3D Gaussian Splatting for Photorealistic Scenes Rendering Open
Accurately rendering scenes with reflective surfaces remains a significant challenge in novel view synthesis, as existing methods like Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) often misinterpret reflections as physica…
View article: PRING: Rethinking Protein-Protein Interaction Prediction from Pairs to Graphs
PRING: Rethinking Protein-Protein Interaction Prediction from Pairs to Graphs Open
Deep learning-based computational methods have achieved promising results in predicting protein-protein interactions (PPIs). However, existing benchmarks predominantly focus on isolated pairwise evaluations, overlooking a model's capabilit…
View article: Large Language Models as Materials Science Adapted Learners
Large Language Models as Materials Science Adapted Learners Open
Materials discovery and design aim to find compositions and structures with desirable properties over highly complex and diverse physical spaces. Traditional solutions, such as high-throughput simulations or machine learning, often rely on…
View article: Model Compression using Progressive Channel Pruning
Model Compression using Progressive Channel Pruning Open
In this work, we propose a simple but effective channel pruning framework called Progressive Channel Pruning (PCP) to accelerate Convolutional Neural Networks (CNNs). In contrast to the existing channel pruning methods that prune channels …
View article: The operational medium-range deterministic weather forecasting can be extended beyond a 10-day lead time
The operational medium-range deterministic weather forecasting can be extended beyond a 10-day lead time Open
Given the complexity of the atmospheric system, current numerical weather prediction models struggle with accurate forecasts. Here we present FengWu, an Artificial-Intelligence-driven global medium-range forecasting system employing multi-…
View article: CSBrain: A Cross-scale Spatiotemporal Brain Foundation Model for EEG Decoding
CSBrain: A Cross-scale Spatiotemporal Brain Foundation Model for EEG Decoding Open
Understanding and decoding brain activity from electroencephalography (EEG) signals is a fundamental challenge in neuroscience and AI, with applications in cognition, emotion recognition, diagnosis, and brain-computer interfaces. While rec…
View article: ShotBench: Expert-Level Cinematic Understanding in Vision-Language Models
ShotBench: Expert-Level Cinematic Understanding in Vision-Language Models Open
Cinematography, the fundamental visual language of film, is essential for conveying narrative, emotion, and aesthetic quality. While recent Vision-Language Models (VLMs) demonstrate strong general visual understanding, their proficiency in…
View article: Dense Video Captioning using Graph-based Sentence Summarization
Dense Video Captioning using Graph-based Sentence Summarization Open
Recently, dense video captioning has made attractive progress in detecting and captioning all events in a long untrimmed video. Despite promising results were achieved, most existing methods do not sufficiently explore the scene evolution …