Lei Bai

Junchao Gong, Tao Han, Kang Chen et al.

Numerical Weather Prediction (NWP) system is an infrastructure that exerts considerable impacts on modern society.Traditional NWP system, however, resolves it by solving complex partial differential equations with a huge computing cluster, resulting in tons of carbon emission. Exploring efficient and eco-friendly solutions for NWP attracts interest from Artificial Intelligence (AI) and earth science communities. To narrow the performance gap between the AI-based methods and physic predictor, this work proposes a new transformer-based NWP framework, termed as WeatherFormer, to model the complex spatio-temporal atmosphere dynamics and empowering the capability of data-driven NWP. WeatherFormer innovatively introduces the space-time factorized transformer blocks to decrease the parameters and memory consumption, in which Position-aware Adaptive Fourier Neural Operator (PAFNO) is proposed for location sensible token mixing. Besides, two data augmentation strategies are utilized to boost the performance and decrease training consumption. Extensive experiments on WeatherBench dataset show WeatherFormer achieves superior performance over existing deep learning methods and further approaches the most advanced physical model.

Chongjun Tu, Peng Ye, Dongzhan Zhou et al.

Multi-Modal Large Language Models (MLLMs) stand out in various tasks but still struggle with hallucinations. While recent training-free mitigation methods mostly introduce additional inference overhead via retrospection strategy and contrastive decoding, we propose attention reallocation (AttnReal) to mitigate hallucinations with nearly zero extra cost. Our approach is motivated by the key observations that, MLLM's unreasonable attention distribution causes features to be dominated by historical output tokens, which further contributes to hallucinated responses because of the distribution gap between different token types. Based on the observations, AttnReal recycles excessive attention from output tokens and reallocates it to visual tokens, which reduces MLLM's reliance on language priors and ensures the decoding process depends more on the visual inputs. More interestingly, we find that, by controlling the intensity of AttnReal, we can achieve a wide-range trade-off between the response faithfulness and overall performance. Comprehensive results from different benchmarks validate the effectiveness of AttnReal across six open-source MLLMs and three decoding strategies.

Fangchen Yu, Haiyuan Wan, Qianjia Cheng et al. · pku, tsinghua

Recently, the physical capabilities of (M)LLMs have garnered increasing attention. However, existing benchmarks for physics suffer from two major gaps: they neither provide systematic and up-to-date coverage of real-world physics competitions such as physics Olympiads, nor enable direct performance comparison with humans. To bridge these gaps, we present HiPhO, the first benchmark dedicated to high school physics Olympiads with human-aligned evaluation. Specifically, HiPhO highlights three key innovations. (1) Comprehensive Data: It compiles 13 latest Olympiad exams from 2024-2025, spanning both international and regional competitions, and covering mixed modalities that encompass problems spanning text-only to diagram-based. (2) Professional Evaluation: We adopt official marking schemes to perform fine-grained grading at both the answer and step level, fully aligned with human examiners to ensure high-quality and domain-specific evaluation. (3) Comparison with Human Contestants: We assign gold, silver, and bronze medals to models based on official medal thresholds, thereby enabling direct comparison between (M)LLMs and human contestants. Our large-scale evaluation of 30 state-of-the-art (M)LLMs shows that: across 13 exams, open-source MLLMs mostly remain at or below the bronze level; open-source LLMs show promising progress with multiple golds; closed-source reasoning MLLMs can achieve 6 to 12 gold medals; and most models still have a significant gap from full marks. These results highlight the performance gap between open-source models and top students, the strong reasoning abilities of closed-source models, and the remaining room for improvement. HiPhO, a human-aligned Olympiad benchmark for multimodal physical reasoning, is open-source at https://github.com/SciYu/HiPhO with a public leaderboard at https://phyarena.github.io/.

Junchao Gong, Lei Bai, Peng Ye et al.

Precipitation nowcasting based on radar data plays a crucial role in extreme weather prediction and has broad implications for disaster management. Despite progresses have been made based on deep learning, two key challenges of precipitation nowcasting are not well-solved: (i) the modeling of complex precipitation system evolutions with different scales, and (ii) accurate forecasts for extreme precipitation. In this work, we propose CasCast, a cascaded framework composed of a deterministic and a probabilistic part to decouple the predictions for mesoscale precipitation distributions and small-scale patterns. Then, we explore training the cascaded framework at the high resolution and conducting the probabilistic modeling in a low dimensional latent space with a frame-wise-guided diffusion transformer for enhancing the optimization of extreme events while reducing computational costs. Extensive experiments on three benchmark radar precipitation datasets show that CasCast achieves competitive performance. Especially, CasCast significantly surpasses the baseline (up to +91.8%) for regional extreme-precipitation nowcasting.

Xiangchao Yan, Shiyang Feng, Jiakang Yuan et al.

Survey paper plays a crucial role in scientific research, especially given the rapid growth of research publications. Recently, researchers have begun using LLMs to automate survey generation for better efficiency. However, the quality gap between LLM-generated surveys and those written by human remains significant, particularly in terms of outline quality and citation accuracy. To close these gaps, we introduce SurveyForge, which first generates the outline by analyzing the logical structure of human-written outlines and referring to the retrieved domain-related articles. Subsequently, leveraging high-quality papers retrieved from memory by our scholar navigation agent, SurveyForge can automatically generate and refine the content of the generated article. Moreover, to achieve a comprehensive evaluation, we construct SurveyBench, which includes 100 human-written survey papers for win-rate comparison and assesses AI-generated survey papers across three dimensions: reference, outline, and content quality. Experiments demonstrate that SurveyForge can outperform previous works such as AutoSurvey.

Yiting Lu, Jiakang Yuan, Zhen Li et al.

We propose OmniCaptioner, a versatile visual captioning framework for generating fine-grained textual descriptions across a wide variety of visual domains. Unlike prior methods limited to specific image types (e.g., natural images or geometric visuals), our framework provides a unified solution for captioning natural images, visual text (e.g., posters, UIs, textbooks), and structured visuals (e.g., documents, tables, charts). By converting low-level pixel information into semantically rich textual representations, our framework bridges the gap between visual and textual modalities. Our results highlight three key advantages: (i) Enhanced Visual Reasoning with LLMs, where long-context captions of visual modalities empower LLMs, particularly the DeepSeek-R1 series, to reason effectively in multimodal scenarios; (ii) Improved Image Generation, where detailed captions improve tasks like text-to-image generation and image transformation; and (iii) Efficient Supervised Fine-Tuning (SFT), which enables faster convergence with less data. We believe the versatility and adaptability of OmniCaptioner can offer a new perspective for bridging the gap between language and visual modalities.

Tianshuo Peng, Mingsheng Li, Jiakang Yuan et al.

Recent advancements in Large Multi-modal Models (LMMs) underscore the importance of scaling by increasing image-text paired data, achieving impressive performance on general tasks. Despite their effectiveness in broad applications, generalist models are primarily trained on web-scale datasets dominated by natural images, resulting in the sacrifice of specialized capabilities for domain-specific tasks that require extensive domain prior knowledge. Moreover, directly integrating expert models tailored for specific domains is challenging due to the representational gap and imbalanced optimization between the generalist model and experts. To address these challenges, we introduce Chimera, a scalable and low-cost multi-modal pipeline designed to boost the ability of existing LMMs with domain-specific experts. Specifically, we design a progressive training strategy to integrate features from expert models into the input of a generalist LMM. To address the imbalanced optimization caused by the well-aligned general visual encoder, we introduce a novel Generalist-Specialist Collaboration Masking (GSCM) mechanism. This results in a versatile model that excels across the chart, table, math, and document domains, achieving state-of-the-art performance on multi-modal reasoning and visual content extraction tasks, both of which are challenging tasks for assessing existing LMMs.

Pumeng Lyu, Tao Tang, Fenghua Ling et al.

Recent studies have shown that deep learning (DL) models can skillfully predict the El Niño-Southern Oscillation (ENSO) forecasts over 1.5 years ahead. However, concerns regarding the reliability of predictions made by DL methods persist, including potential overfitting issues and lack of interpretability. Here, we propose ResoNet, a DL model that combines convolutional neural network (CNN) and Transformer architectures. This hybrid architecture design enables our model to adequately capture local SSTA as well as long-range inter-basin interactions across oceans. We show that ResoNet can robustly predict ESNO at lead times between 19 and 26 months, thus outperforming existing approaches in terms of the forecast horizon. According to an explainability method applied to ResoNet predictions of El Niño and La Niña events from 1- to 18-month lead, we find that it predicts the Niño3.4 index based on multiple physically reasonable mechanisms, such as the Recharge Oscillator concept, Seasonal Footprint Mechanism, and Indian Ocean capacitor effect. Moreover, we demonstrate that for the first time, the asymmetry between El Niño and La Niña development can be captured by ResoNet. Our results could help alleviate skepticism about applying DL models for ENSO prediction and encourage more attempts to discover and predict climate phenomena using AI methods.

Wanghan Xu, Xiaoyu Yue, Zidong Wang et al.

Generative models serve as powerful tools for modeling the real world, with mainstream diffusion models, particularly those based on the latent diffusion model paradigm, achieving remarkable progress across various tasks, such as image and video synthesis. Latent diffusion models are typically trained using Variational Autoencoders (VAEs), interacting with VAE latents rather than the real samples. While this generative paradigm speeds up training and inference, the quality of the generated outputs is limited by the latents' quality. Traditional VAE latents are often seen as spatial compression in pixel space and lack explicit semantic representations, which are essential for modeling the real world. In this paper, we introduce ReaLS (Representation-Aligned Latent Space), which integrates semantic priors to improve generation performance. Extensive experiments show that fundamental DiT and SiT trained on ReaLS can achieve a 15% improvement in FID metric. Furthermore, the enhanced semantic latent space enables more perceptual downstream tasks, such as segmentation and depth estimation.

Lintao Wang, Encheng Su, Jiaqi Liu et al.

Physics problem-solving is a challenging domain for large AI models, requiring integration of conceptual understanding, mathematical reasoning, and interpretation of physical diagrams. Current evaluation methodologies show notable limitations in capturing the breadth and complexity of undergraduate-level physics, underscoring the need for more rigorous assessments. To this end, we present PhysUniBench, a large-scale multimodal benchmark designed to evaluate and improve the reasoning capabilities of multimodal large language models (MLLMs) specifically on undergraduate-level physics problems. PhysUniBench consists of 3,304 physics questions spanning 8 major sub-disciplines of physics, each accompanied by one visual diagrams. The benchmark includes both open-ended and multiple-choice questions, systematically curated and difficulty-rated through an iterative model-in-the-loop process. The benchmark's construction involved a rigorous multi-stage process, including multiple roll-outs, expert-level evaluation, automated filtering of easily solved problems, and a nuanced difficulty grading system with five levels. Through extensive experiments, we observe that current state-of-the-art models encounter substantial challenges in physics reasoning. For example, GPT-4o mini achieves only about 34.2% accuracy in the proposed PhysUniBench. These results highlight that current MLLMs struggle with advanced physics reasoning, especially on multi-step problems and those requiring precise diagram interpretation. By providing a broad and rigorous assessment tool, PhysUniBench aims to drive progress in AI for Science, encouraging the development of models with stronger physical reasoning, problem-solving skills, and multimodal understanding. The benchmark and evaluation scripts are available at https://prismax-team.github.io/PhysUniBenchmark/.

Xinyu Wang, Lei Liu, Kang Chen et al.

Tropical cyclone (TC) intensity forecasting is crucial for early disaster warning and emergency decision-making. Numerous researchers have explored deep-learning methods to address computational and post-processing issues in operational forecasting. Regrettably, they exhibit subpar long-term forecasting capabilities. We use two strategies to enhance long-term forecasting. (1) By enhancing the matching between TC intensity and spatial information, we can improve long-term forecasting performance. (2) Incorporating physical knowledge and physical constraints can help mitigate the accumulation of forecasting errors. To achieve the above strategies, we propose the VQLTI framework. VQLTI transfers the TC intensity information to a discrete latent space while retaining the spatial information differences, using large-scale spatial meteorological data as conditions. Furthermore, we leverage the forecast from the weather prediction model FengWu to provide additional physical knowledge for VQLTI. Additionally, we calculate the potential intensity (PI) to impose physical constraints on the latent variables. In the global long-term TC intensity forecasting, VQLTI achieves state-of-the-art results for the 24h to 120h, with the MSW (Maximum Sustained Wind) forecast error reduced by 35.65%-42.51% compared to ECMWF-IFS.