Large language models (LLMs) are trained on enormous documents that contain extensive world knowledge. However, it is still not well-understood how knowledge is acquired via autoregressive pre-training. This lack of understanding greatly hinders effective knowledge learning, especially for continued pretraining on up-to-date information, as this evolving information often lacks diverse repetitions like foundational knowledge. In this paper, we focus on understanding and improving LLM knowledge learning. We found and verified that knowledge learning for LLMs can be deemed as an implicit supervised task hidden in the autoregressive pre-training objective. Our findings suggest that knowledge learning for LLMs would benefit from methods designed to improve generalization ability for supervised tasks. Based on our analysis, we propose the formatting-based data augmentation to grow in-distribution samples, which does not present the risk of altering the facts embedded in documents as text paraphrasing. We also introduce sharpness-aware minimization as an effective optimization algorithm to better improve generalization. Moreover, our analysis and method can be readily extended to instruction tuning. Extensive experiment results validate our findings and demonstrate our methods' effectiveness in both continued pre-training and instruction tuning. This paper offers new perspectives and insights to interpret and design effective strategies for LLM knowledge learning.
We build a state-of-the-art dynamic model of private asset allocation that considers five key features of private asset markets: (1) the illiquid nature of private assets, (2) timing lags between capital commitments, capital calls, and eventual distributions, (3) time-varying business cycle conditions, (4) serial correlation in observed private asset returns, and (5) regulatory constraints on certain institutional investors' portfolio choices. We use cutting-edge machine learning methods to quantify the optimal investment policies over the life cycle of a fund. Moreover, our model offers regulators a tool for precisely quantifying the trade-offs when setting risk-based capital charges.
Realistic and interactive surgical simulation has the potential to facilitate crucial applications, such as medical professional training and autonomous surgical agent training. In the natural visual domain, world models have enabled action-controlled data generation, demonstrating the potential to train autonomous agents in interactive simulated environments when large-scale real data acquisition is infeasible. However, such works in the surgical domain have been limited to simplified computer simulations, and lack realism. Furthermore, existing literature in world models has predominantly dealt with action-labeled data, limiting their applicability to real-world surgical data, where obtaining action annotation is prohibitively expensive. Inspired by the recent success of Genie in leveraging unlabeled video game data to infer latent actions and enable action-controlled data generation, we propose the first surgical vision world model. The proposed model can generate action-controllable surgical data and the architecture design is verified with extensive experiments on the unlabeled SurgToolLoc-2022 dataset. Codes and implementation details are available at https://github.com/bhattarailab/Surgical-Vision-World-Model
We introduce a new dynamic factor correlation model with a novel variation-free parametrization of factor loadings. The model is applicable to high dimensions and can accommodate time-varying correlations, heterogeneous heavy-tailed distributions, and dependent idiosyncratic shocks, such as those observed in returns on stocks in the same subindustry. We apply the model to a "small universe" with 12 asset returns and to a "large universe" with 323 asset returns. The former facilitates a comprehensive empirical analysis and comparisons and the latter demonstrates the flexibility and scalability of the model.
We introduce DAMCC (Deep Autoregressive Model for Dynamic Combinatorial Complexes), the first deep learning model designed to generate dynamic combinatorial complexes (CCs). Unlike traditional graph-based models, CCs capture higher-order interactions, making them ideal for representing social networks, biological systems, and evolving infrastructures. While existing models primarily focus on static graphs, DAMCC addresses the challenge of modeling temporal dynamics and higher-order structures in dynamic networks. DAMCC employs an autoregressive framework to predict the evolution of CCs over time. Through comprehensive experiments on real-world and synthetic datasets, we demonstrate its ability to capture both temporal and higher-order dependencies. As the first model of its kind, DAMCC lays the foundation for future advancements in dynamic combinatorial complex modeling, with opportunities for improved scalability and efficiency on larger networks.
Hypergraph neural networks (HGNNs) effectively model complex high-order relationships in domains like protein interactions and social networks by connecting multiple vertices through hyperedges, enhancing modeling capabilities, and reducing information loss. Developing foundation models for hypergraphs is challenging due to their distinct data, which includes both vertex features and intricate structural information. We present Hyper-FM, a Hypergraph Foundation Model for multi-domain knowledge extraction, featuring Hierarchical High-Order Neighbor Guided Vertex Knowledge Embedding for vertex feature representation and Hierarchical Multi-Hypergraph Guided Structural Knowledge Extraction for structural information. Additionally, we curate 10 text-attributed hypergraph datasets to advance research between HGNNs and LLMs. Experiments on these datasets show that Hyper-FM outperforms baseline methods by approximately 13.3\%, validating our approach. Furthermore, we propose the first scaling law for hypergraph foundation models, demonstrating that increasing domain diversity significantly enhances performance, unlike merely augmenting vertex and hyperedge counts. This underscores the critical role of domain diversity in scaling hypergraph models.
Differentiable environments have heralded new possibilities for learning control policies by offering rich differentiable information that facilitates gradient-based methods. In comparison to prevailing model-free reinforcement learning approaches, model-based reinforcement learning (MBRL) methods exhibit the potential to effectively harness the power of differentiable information for recovering the underlying physical dynamics. However, this presents two primary challenges: effectively utilizing differentiable information to 1) construct models with more accurate dynamic prediction and 2) enhance the stability of policy training. In this paper, we propose a Differentiable Information Enhanced MBRL method, MB-MIX, to address both challenges. Firstly, we adopt a Sobolev model training approach that penalizes incorrect model gradient outputs, enhancing prediction accuracy and yielding more precise models that faithfully capture system dynamics. Secondly, we introduce mixing lengths of truncated learning windows to reduce the variance in policy gradient estimation, resulting in improved stability during policy learning. To validate the effectiveness of our approach in differentiable environments, we provide theoretical analysis and empirical results. Notably, our approach outperforms previous model-based and model-free methods, in multiple challenging tasks involving controllable rigid robots such as humanoid robots' motion control and deformable object manipulation.
The integration of deep generative networks into generating Computer-Aided Design (CAD) models has garnered increasing attention over recent years. Traditional methods often rely on discrete sequences of parametric line/curve segments to represent sketches. Differently, we introduce RECAD, a novel framework that generates Raster sketches and 3D Extrusions for CAD models. Representing sketches as raster images offers several advantages over discrete sequences: 1) it breaks the limitations on the types and numbers of lines/curves, providing enhanced geometric representation capabilities; 2) it enables interpolation within a continuous latent space; and 3) it allows for more intuitive user control over the output. Technically, RECAD employs two diffusion networks: the first network generates extrusion boxes conditioned on the number and types of extrusions, while the second network produces sketch images conditioned on these extrusion boxes. By combining these two networks, RECAD effectively generates sketch-and-extrude CAD models, offering a more robust and intuitive approach to CAD model generation. Experimental results indicate that RECAD achieves strong performance in unconditional generation, while also demonstrating effectiveness in conditional generation and output editing.
A unified video and action model holds significant promise for robotics, where videos provide rich scene information for action prediction, and actions provide dynamics information for video prediction. However, effectively combining video generation and action prediction remains challenging, and current video generation-based methods struggle to match the performance of direct policy learning in action accuracy and inference speed. To bridge this gap, we introduce the Unified Video Action model (UVA), which jointly optimizes video and action predictions to achieve both high accuracy and efficient action inference. The key lies in learning a joint video-action latent representation and decoupling video-action decoding. The joint latent representation bridges the visual and action domains, effectively modeling the relationship between video and action sequences. Meanwhile, the decoupled decoding, powered by two lightweight diffusion heads, enables high-speed action inference by bypassing video generation during inference. Such a unified framework further enables versatile functionality through masked input training. By selectively masking actions or videos, a single model can tackle diverse tasks beyond policy learning, such as forward and inverse dynamics modeling and video generation. Via an extensive set of experiments, we demonstrate that UVA can serve as a general-purpose solution for a wide range of robotics tasks, such as policy learning, forward/inverse dynamics and video observation prediction, without compromising performance compared to methods tailored for specific applications. Results are best viewed on https://unified-video-action-model.github.io/.
Surrogate-assisted evolutionary algorithms (SAEAs) have been proposed to solve expensive optimization problems. Although SAEAs use surrogate models that approximate the evaluations of solutions using machine learning techniques, prior research has not adequately investigated the impact of surrogate model accuracy on search performance and model management strategy in SAEAs. This study analyzes how surrogate model accuracy affects search performance and model management strategies. For this purpose, we construct a pseudo-surrogate model with adjustable prediction accuracy to ensure fair comparisons across different model management strategies. We compared three model management strategies: (1) pre-selection (PS), (2) individual-based (IB), and (3) generation-based (GB) on standard benchmark problems with a baseline model that does not use surrogates. The experimental results reveal that a higher surrogate model accuracy improves the search performance. However, the impact varies according to the strategy used. Specifically, PS demonstrates a clear trend of improved performance as the estimation accuracy increases, whereas IB and GB exhibit robust performance when the accuracy surpasses a certain threshold. In model strategy comparisons, GB exhibits superior performance across a broad range of prediction accuracies, IB outperforms it at lower accuracies, and PS outperforms it at higher accuracies. The findings of this study clarify guidelines for selecting appropriate model management strategies based on the surrogate model accuracy.
Brain-inspired spiking neural networks (SNNs) have garnered significant research attention in algorithm design and perception applications. However, their potential in the decision-making domain, particularly in model-based reinforcement learning, remains underexplored. The difficulty lies in the need for spiking neurons with long-term temporal memory capabilities, as well as network optimization that can integrate and learn information for accurate predictions. The dynamic dendritic information integration mechanism of biological neurons brings us valuable insights for addressing these challenges. In this study, we propose a multi-compartment neuron model capable of nonlinearly integrating information from multiple dendritic sources to dynamically process long sequential inputs. Based on this model, we construct a Spiking World Model (Spiking-WM), to enable model-based deep reinforcement learning (DRL) with SNNs. We evaluated our model using the DeepMind Control Suite, demonstrating that Spiking-WM outperforms existing SNN-based models and achieves performance comparable to artificial neural network (ANN)-based world models employing Gated Recurrent Units (GRUs). Furthermore, we assess the long-term memory capabilities of the proposed model in speech datasets, including SHD, TIMIT, and LibriSpeech 100h, showing that our multi-compartment neuron model surpasses other SNN-based architectures in processing long sequences. Our findings underscore the critical role of dendritic information integration in shaping neuronal function, emphasizing the importance of cooperative dendritic processing in enhancing neural computation.
We characterize regularity (Block & Marschak, 1960) within a novel stochastic model: the General Threshold Luce model [GTLM]. We apply our results to study choice overload, identified by regularity violations that impose a welfare cost on the decision-maker. Generalizing our characterization results, we identify necessary and sufficient conditions for choice overload within GTLMs and, in doing so, disentangle two well-known causes: low discriminatory power (Frick, 2016) and limited attention (Lleras et al., 2017).
Entanglement is a fundamental feature of quantum mechanics, playing a crucial role in quantum information processing. However, classifying entangled states, particularly in the mixed-state regime, remains a challenging problem, especially as system dimensions increase. In this work, we focus on bipartite quantum states and present a data-driven approach to entanglement classification using transformer-based neural networks. Our dataset consists of a diverse set of bipartite states, including pure separable states, Werner entangled states, general entangled states, and maximally entangled states. We pretrain the transformer in an unsupervised fashion by masking elements of vectorized Hermitian matrix representations of quantum states, allowing the model to learn structural properties of quantum density matrices. This approach enables the model to generalize entanglement characteristics across different classes of states. Once trained, our method achieves near-perfect classification accuracy, effectively distinguishing between separable and entangled states. Compared to previous Machine Learning, our method successfully adapts transformers for quantum state analysis, demonstrating their ability to systematically identify entanglement in bipartite systems. These results highlight the potential of modern machine learning techniques in automating entanglement detection and classification, bridging the gap between quantum information theory and artificial intelligence.
Model merging integrates the parameters of multiple models into a unified model, combining their diverse capabilities. Existing model merging methods are often constrained by fixed parameter merging ratios. In this study, we propose Mixup Model Merge (M$^3$), an innovative approach inspired by the Mixup data augmentation technique. This method merges the parameters of two large language models (LLMs) by randomly generating linear interpolation ratios, allowing for a more flexible and comprehensive exploration of the parameter space. Extensive experiments demonstrate the superiority of our proposed M$^3$ method in merging fine-tuned LLMs: (1) it significantly improves performance across multiple tasks, (2) it enhances LLMs' out-of-distribution (OOD) robustness and adversarial robustness, (3) it achieves superior results when combined with sparsification techniques such as DARE, and (4) it offers a simple yet efficient solution that does not require additional computational resources. In conclusion, M$^3$ is a simple yet effective model merging method that significantly enhances the performance of the merged model by randomly generating contribution ratios for two fine-tuned LLMs. The code is available at https://github.com/MLGroupJLU/MixupModelMerge.
Reliable reservoir volume estimates are crucial for water resource management, especially in arid and semi-arid regions. The present study investigates applying three machine learning regression techniques - Gradient Boosting, Random Forest, and ElasticNet to predict key dam performance characteristics of the Loskop Dam in South Africa. The models were trained and validated on a dataset comprising geospatial elevation measurements paired with corresponding reservoir supply capacity values. The best-performing approach was a threshold-based blended model that combined random forest for higher volumes with Ridge regression for lower volumes. This model achieved an RMSE of 4.88 MCM and an R2 of 0.99. These findings highlight the ability of ensemble learning techniques to capture complex relationships in dam datasets and underscore their practical utility for reliable dam performance modelling in real-world water resource management scenarios.
Hyperspectral imaging provides detailed spectral information, offering significant potential for monitoring greenhouse gases like CH4 and NO2. However, its application is constrained by limited spatial coverage and infrequent revisit times. In contrast, multispectral imaging delivers broader spatial and temporal coverage but lacks the spectral granularity required for precise GHG detection. To address these challenges, this study proposes Spectral and Spatial-Spectral transformer models that reconstruct hyperspectral data from multispectral inputs. The models in this paper are pretrained on EnMAP and EMIT datasets and fine-tuned on spatio-temporally aligned (Sentinel-2, EnMAP) and (HLS-S30, EMIT) image pairs respectively. Our model has the potential to enhance atmospheric monitoring by combining the strengths of hyperspectral and multispectral imaging systems.
State-of-the-art large language and vision models are trained over trillions of tokens that are aggregated from a large variety of sources. As training data collections grow, manually managing the samples becomes time-consuming, tedious, and prone to errors. Yet recent research shows that the data mixture and the order in which samples are visited during training can significantly influence model accuracy. We build and present Mixtera, a data plane for foundation model training that enables users to declaratively express which data samples should be used in which proportion and in which order during training. Mixtera is a centralized, read-only layer that is deployed on top of existing training data collections and can be declaratively queried. It operates independently of the filesystem structure and supports mixtures across arbitrary properties (e.g., language, source dataset) as well as dynamic adjustment of the mixture based on model feedback. We experimentally evaluate Mixtera and show that our implementation does not bottleneck training and scales to 256 GH200 superchips. We demonstrate how Mixtera supports recent advancements in mixing strategies by implementing the proposed Adaptive Data Optimization (ADO) algorithm in the system and evaluating its performance impact. We also explore the role of mixtures for vision-language models.
AI-agility, with which an organization can be quickly adapted to its business priorities, is desired even for the development and operations of generative AI (GenAI) applications. Especially in this paper, we discuss so-called GenAI Model-agility, which we define as the readiness to be flexibly adapted to base foundation models as diverse as the model providers and versions. First, for handling issues specific to generative AI, we first define a methodology of GenAI application development and operations, as GenAIOps, to identify the problem of application quality degradation caused by changes to the underlying foundation models. We study prompt tuning technologies, which look promising to address this problem, and discuss their effectiveness and limitations through case studies using existing tools.
Algebraic model counting unifies many inference tasks on logic formulas by exploiting semirings. Rather than focusing on inference, we consider learning, especially in statistical-relational and neurosymbolic AI, which combine logical, probabilistic and neural representations. Concretely, we show that the very same semiring perspective of algebraic model counting also applies to learning. This allows us to unify various learning algorithms by generalizing gradients and backpropagation to different semirings. Furthermore, we show how cancellation and ordering properties of a semiring can be exploited for more memory-efficient backpropagation. This allows us to obtain some interesting variations of state-of-the-art gradient-based optimisation methods for probabilistic logical models. We also discuss why algebraic model counting on tractable circuits does not lead to more efficient second-order optimization. Empirically, our algebraic backpropagation exhibits considerable speed-ups as compared to existing approaches.
Electroencephalography provides a non-invasive window into brain activity, offering valuable insights for neurological research, brain-computer interfaces, and clinical diagnostics. However, the development of robust machine learning models for EEG analysis is hindered by the scarcity of large-scale, well-annotated datasets and the inherent variability of EEG signals across subjects and recording conditions. Inspired by the success of foundation models in natural language processing and computer vision, we propose the Large Cognition Model-a transformer-based foundation model designed to generalize across diverse EEG datasets and downstream tasks. Unlike traditional approaches, our proposed transformer-based architecture demonstrates strong generalization capabilities across datasets and tasks, even without pretraining, surpassing some existing EEG universal models on specific downstream applications. LCM leverages large-scale self-supervised learning techniques to capture universal EEG representations, enabling efficient fine-tuning for applications such as cognitive state decoding, disease classification, and neurofeedback systems. We introduce a novel architecture that integrates temporal and spectral attention mechanisms, optimizing the model's ability to extract meaningful features from raw EEG signals. Extensive evaluations demonstrate that LCM outperforms state-of-the-art approaches across multiple EEG benchmarks, exhibiting strong cross-subject and cross-task generalization. Our findings highlight the potential of pretrained EEG foundation models to accelerate advancements in neuroscience, personalized medicine, and BCI technology.