Federated Learning (FL) trains machine learning models on edge devices with distributed data. However, the computational and memory limitations of these devices restrict the training of large models using FL. Split Federated Learning (SFL) addresses this challenge by distributing the model across the device and server, but it introduces a tightly coupled data flow, leading to computational bottlenecks and high communication costs. We propose EMO as a solution to enable the training of large models in FL while mitigating the challenges of SFL. EMO introduces Edge Model Overlay(s) between the device and server, enabling the creation of a larger ensemble model without modifying the FL workflow. The key innovation in EMO is Augmented Federated Learning (AFL), which builds an ensemble model by connecting the original (smaller) FL model with model(s) trained in the overlay(s) to facilitate horizontal or vertical scaling. This is accomplished through three key modules: a hierarchical activation replay cache to decouple AFL from FL, a convergence-aware communication controller to optimize communication overhead, and an ensemble inference module. Evaluations on a real-world prototype show that EMO improves accuracy by up to 17.77% compared to FL, and reduces communication costs by up to 7.17x and decreases training time by up to 6.9x compared to SFL.
Model calibration is essential for ensuring that the predictions of deep neural networks accurately reflect true probabilities in real-world classification tasks. However, deep networks often produce over-confident or under-confident predictions, leading to miscalibration. Various methods have been proposed to address this issue by designing effective loss functions for calibration, such as focal loss. In this paper, we analyze its effectiveness and provide a unified loss framework of focal loss and its variants, where we mainly attribute their superiority in model calibration to the loss weighting factor that estimates sample-wise uncertainty. Based on our analysis, existing loss functions fail to achieve optimal calibration performance due to two main issues: including misalignment during optimization and insufficient precision in uncertainty estimation. Specifically, focal loss cannot align sample uncertainty with gradient scaling and the single logit cannot indicate the uncertainty. To address these issues, we reformulate the optimization from the perspective of gradients, which focuses on uncertain samples. Meanwhile, we propose using the Brier Score as the loss weight factor, which provides a more accurate uncertainty estimation via all the logits. Extensive experiments on various models and datasets demonstrate that our method achieves state-of-the-art (SOTA) performance.
The rapid expansion of multi-source satellite imagery drives innovation in Earth observation, opening unprecedented opportunities for Remote Sensing Foundation Models to harness diverse data. However, many existing models remain constrained by fixed spatial resolutions and patch sizes, limiting their ability to fully exploit the heterogeneous spatial characteristics inherent in satellite imagery. To address these challenges, we propose FlexiMo, a flexible remote sensing foundation model that endows the pre-trained model with the flexibility to adapt to arbitrary spatial resolutions. Central to FlexiMo is a spatial resolution-aware module that employs a parameter-free alignment embedding mechanism to dynamically recalibrate patch embeddings based on the input image's resolution and dimensions. This design not only preserves critical token characteristics and ensures multi-scale feature fidelity but also enables efficient feature extraction without requiring modifications to the underlying network architecture. In addition, FlexiMo incorporates a lightweight channel adaptation module that leverages prior spectral information from sensors. This mechanism allows the model to process images with varying numbers of channels while maintaining the data's intrinsic physical properties. Extensive experiments on diverse multimodal, multi-resolution, and multi-scale datasets demonstrate that FlexiMo significantly enhances model generalization and robustness. In particular, our method achieves outstanding performance across a range of downstream tasks, including scene classification, land cover classification, urban building segmentation, and cloud detection. By enabling parameter-efficient and physically consistent adaptation, FlexiMo paves the way for more adaptable and effective foundation models in real-world remote sensing applications.
Early detection, accurate segmentation, classification and tracking of polyps during colonoscopy are critical for preventing colorectal cancer. Many existing deep-learning-based methods for analyzing colonoscopic videos either require task-specific fine-tuning, lack tracking capabilities, or rely on domain-specific pre-training. In this paper, we introduce \textit{PolypSegTrack}, a novel foundation model that jointly addresses polyp detection, segmentation, classification and unsupervised tracking in colonoscopic videos. Our approach leverages a novel conditional mask loss, enabling flexible training across datasets with either pixel-level segmentation masks or bounding box annotations, allowing us to bypass task-specific fine-tuning. Our unsupervised tracking module reliably associates polyp instances across frames using object queries, without relying on any heuristics. We leverage a robust vision foundation model backbone that is pre-trained unsupervisedly on natural images, thereby removing the need for domain-specific pre-training. Extensive experiments on multiple polyp benchmarks demonstrate that our method significantly outperforms existing state-of-the-art approaches in detection, segmentation, classification, and tracking.
Reliability of LLMs is questionable even as they get better at more tasks. A wider adoption of LLMs is contingent on whether they are usably factual. And if they are not, on whether they can properly calibrate their confidence in their responses. This work focuses on utilizing the multilingual knowledge of an LLM to inform its decision to abstain or answer when prompted. We develop a multilingual pipeline to calibrate the model's confidence and let it abstain when uncertain. We run several multilingual models through the pipeline to profile them across different languages. We find that the performance of the pipeline varies by model and language, but that in general they benefit from it. This is evidenced by the accuracy improvement of $71.2\%$ for Bengali over a baseline performance without the pipeline. Even a high-resource language like English sees a $15.5\%$ improvement. These results hint at possible further improvements.
Reinforcement learning from human feedback (RLHF) is a critical technique for training large language models. However, conventional reward models based on the Bradley-Terry model (BTRM) often suffer from overconfidence when faced with inconsistent labels or out-of-distribution samples, leading to reward hacking, where the policy model blindly optimizes for proxy rewards while degrading true performance. This paper proposes the Probabilistic Uncertain Reward Model (PURM), which generalizes the Bradley-Terry model to learn the reward distributions that emerged from the preference data. We theoretically derive the loss function of PURM and introduce a novel method that uses the overlap between distributions to quantify uncertainty. Empirical results show that PURM outperforms existing methods with more accurate reward and sound uncertainty estimations, and sustains effective learning for more optimization steps and obtain higher maximum win rate in RLHF. The data and code of this paper are released at https://anonymous.4open.science/r/Probabilistic-Uncertain-Reward-Model/
Model merging has emerged as a promising approach for unifying independently fine-tuned models into an integrated framework, significantly enhancing computational efficiency in multi-task learning. Recently, several SVD-based techniques have been introduced to exploit low-rank structures for enhanced merging, but their reliance on such manually designed rank selection often leads to cross-task interference and suboptimal performance. In this paper, we propose AdaRank, a novel model merging framework that adaptively selects the most beneficial singular directions of task vectors to merge multiple models. We empirically show that the dominant singular components of task vectors can cause critical interference with other tasks, and that naive truncation across tasks and layers degrades performance. In contrast, AdaRank dynamically prunes the singular components that cause interference and offers an optimal amount of information to each task vector by learning to prune ranks during test-time via entropy minimization. Our analysis demonstrates that such method mitigates detrimental overlaps among tasks, while empirical results show that AdaRank consistently achieves state-of-the-art performance with various backbones and number of tasks, reducing the performance gap between fine-tuned models to nearly 1%.
Reconstructing dynamic assets from video data is central to many in computer vision and graphics tasks. Existing 4D reconstruction approaches are limited by category-specific models or slow optimization-based methods. Inspired by the recent Large Reconstruction Model (LRM), we present the Large Interpolation Model (LIM), a transformer-based feed-forward solution, guided by a novel causal consistency loss, for interpolating implicit 3D representations across time. Given implicit 3D representations at times $t_0$ and $t_1$, LIM produces a deformed shape at any continuous time $t\in[t_0,t_1]$, delivering high-quality interpolated frames in seconds. Furthermore, LIM allows explicit mesh tracking across time, producing a consistently uv-textured mesh sequence ready for integration into existing production pipelines. We also use LIM, in conjunction with a diffusion-based multiview generator, to produce dynamic 4D reconstructions from monocular videos. We evaluate LIM on various dynamic datasets, benchmarking against image-space interpolation methods (e.g., FiLM) and direct triplane linear interpolation, and demonstrate clear advantages. In summary, LIM is the first feed-forward model capable of high-speed tracked 4D asset reconstruction across diverse categories.
Curse of Dimensionality is an unavoidable challenge in statistical probability models, yet diffusion models seem to overcome this limitation, achieving impressive results in high-dimensional data generation. Diffusion models assume that they can learn the statistical properties of the underlying probability distribution, enabling sampling from this distribution to generate realistic samples. But is this really how they work? To address this question, this paper conducts a detailed analysis of the objective function and inference methods of diffusion models, leading to several important conclusions that help answer the above question: 1) In high-dimensional sparse scenarios, the target of the objective function fitting degrades from a weighted sum of multiple samples to a single sample. 2) The mainstream inference methods can all be represented within a simple unified framework, without requiring statistical concepts such as Markov chains and SDE, while aligning with the degraded objective function. 3) Guided by this simple framework, more efficient inference methods can be discovered.
Accurately quantifying a large language model's (LLM) predictive uncertainty is crucial for judging the reliability of its answers. While most existing research focuses on short, directly answerable questions with closed-form outputs (e.g., multiple-choice), involving intermediate reasoning steps in LLM responses is increasingly important. This added complexity complicates uncertainty quantification (UQ) because the probabilities assigned to answer tokens are conditioned on a vast space of preceding reasoning tokens. Direct marginalization is infeasible, and the dependency inflates probability estimates, causing overconfidence in UQ. To address this, we propose UQAC, an efficient method that narrows the reasoning space to a tractable size for marginalization. UQAC iteratively constructs an "attention chain" of tokens deemed "semantically crucial" to the final answer via a backtracking procedure. Starting from the answer tokens, it uses attention weights to identify the most influential predecessors, then iterates this process until reaching the input tokens. Similarity filtering and probability thresholding further refine the resulting chain, allowing us to approximate the marginal probabilities of the answer tokens, which serve as the LLM's confidence. We validate UQAC on multiple reasoning benchmarks with advanced open-source LLMs, demonstrating that it consistently delivers reliable UQ estimates with high computational efficiency.
Adapting Large Language Models (LLMs) to downstream tasks using Reinforcement Learning (RL) has proven to be an effective approach. However, LLMs do not inherently define the structure of an agent for RL training, particularly in terms of defining the action space. This paper studies learning a compact latent action space to enhance the controllability and exploration of RL for LLMs. We propose Controlling Large Language Models with Latent Actions (CoLA), a framework that integrates a latent action space into pre-trained LLMs. We apply CoLA to the Llama-3.1-8B model. Our experiments demonstrate that, compared to RL with token-level actions, CoLA's latent action enables greater semantic diversity in text generation. For enhancing downstream tasks, we show that CoLA with RL achieves a score of 42.4 on the math500 benchmark, surpassing the baseline score of 38.2, and reaches 68.2 when augmented with a Monte Carlo Tree Search variant. Furthermore, CoLA with RL consistently improves performance on agent-based tasks without degrading the pre-trained LLM's capabilities, unlike the baseline. Finally, CoLA reduces computation time by half in tasks involving enhanced thinking prompts for LLMs by RL. These results highlight CoLA's potential to advance RL-based adaptation of LLMs for downstream applications.
Despite the significant advancements in general image segmentation achieved by large-scale pre-trained foundation models (such as Meta's Segment Any-thing Model (SAM) series and DINOv2), their performance in specialized fields remains limited by two critical issues: the excessive training costs due to large model parameters, and the insufficient ability to represent specific domain characteristics. This paper proposes a multi-scale feature collabora-tion framework guided by DINOv2 for SAM2, with core innovations in three aspects: (1) Establishing a feature collaboration mechanism between DINOv2 and SAM2 backbones, where high-dimensional semantic features extracted by the self-supervised model guide multi-scale feature fusion; (2) Designing lightweight adapter modules and cross-modal, cross-layer feature fusion units to inject cross-domain knowledge while freezing the base model parameters; (3) Constructing a U-shaped network structure based on U-net, which utilizes attention mechanisms to achieve adaptive aggregation decoding of multi-granularity features. This framework surpasses existing state-of-the-art meth-ods in downstream tasks such as camouflage target detection and salient ob-ject detection, without requiring costly training processes. It provides a tech-nical pathway for efficient deployment of visual image segmentation, demon-strating significant application value in a wide range of downstream tasks and specialized fields within image segmentation.Project page: https://github.com/CheneyXuYiMin/SAM2DINO-Seg
Reconstructing dynamic 3D scenes (i.e., 4D geometry) from monocular video is an important yet challenging problem. Conventional multiview geometry-based approaches often struggle with dynamic motion, whereas recent learning-based methods either require specialized 4D representation or sophisticated optimization. In this paper, we present Sora3R, a novel framework that taps into the rich spatiotemporal priors of large-scale video diffusion models to directly infer 4D pointmaps from casual videos. Sora3R follows a two-stage pipeline: (1) we adapt a pointmap VAE from a pretrained video VAE, ensuring compatibility between the geometry and video latent spaces; (2) we finetune a diffusion backbone in combined video and pointmap latent space to generate coherent 4D pointmaps for every frame. Sora3R operates in a fully feedforward manner, requiring no external modules (e.g., depth, optical flow, or segmentation) or iterative global alignment. Extensive experiments demonstrate that Sora3R reliably recovers both camera poses and detailed scene geometry, achieving performance on par with state-of-the-art methods for dynamic 4D reconstruction across diverse scenarios.
In traditional innovation practices, concept and IP generation are often iteratively integrated. Both processes demand an intricate understanding of advanced technical domain knowledge. Existing large language models (LLMs), while possessing massive pre-trained knowledge, often fall short in the innovative concept generation due to a lack of specialized knowledge necessary for the generation. To bridge this critical gap, we propose a novel knowledge finetuning (KFT) framework to endow LLM-based AI with the ability to autonomously mine, understand, and apply domain-specific knowledge and concepts for invention generation, i.e., concept and patent generation together. Our proposed PatentGPT integrates knowledge injection pre-training (KPT), domain-specific supervised finetuning (SFT), and reinforcement learning from human feedback (RLHF). Extensive evaluation shows that PatentGPT significantly outperforms the state-of-the-art models on patent-related benchmark tests. Our method not only provides new insights into data-driven innovation but also paves a new path to fine-tune LLMs for applications in the context of technology. We also discuss the managerial and policy implications of AI-generating inventions in the future.
The success of large language models has garnered widespread attention for model merging techniques, especially training-free methods which combine model capabilities within the parameter space. However, two challenges remain: (1) uniform treatment of all parameters leads to performance degradation; (2) search-based algorithms are often inefficient. In this paper, we present an innovative framework termed Reinforced Model Merging (RMM), which encompasses an environment and agent tailored for merging tasks. These components interact to execute layer-wise merging actions, aiming to search the optimal merging architecture. Notably, RMM operates without any gradient computations on the original models, rendering it feasible for edge devices. Furthermore, by utilizing data subsets during the evaluation process, we addressed the bottleneck in the reward feedback phase, thereby accelerating RMM by up to 100 times. Extensive experiments demonstrate that RMM achieves state-of-the-art performance across various vision and NLP datasets and effectively overcomes the limitations of the existing baseline methods. Our code is available at https://github.com/WuDiHJQ/Reinforced-Model-Merging.
Mitigating Trojans in Large Language Models (LLMs) is one of many tasks where alignment data is LLM specific, as different LLMs have different Trojan triggers and trigger behaviors to be removed. In this paper, we introduce TeleLoRA (Teleporting Low-Rank Adaptation), a novel framework that synergizes model-specific alignment data across multiple LLMs to enable zero-shot Trojan mitigation on unseen LLMs without alignment data. TeleLoRA learns a unified generator of LoRA adapter weights by leveraging local activation information across multiple LLMs. This generator is designed to be permutation symmetric to generalize across models with different architectures and sizes. We optimize the model design for memory efficiency, making it feasible to learn with large-scale LLMs with minimal computational resources. Experiments on LLM Trojan mitigation benchmarks demonstrate that TeleLoRA effectively reduces attack success rates while preserving the benign performance of the models.
Modern LLMs struggle with efficient updates, as each new pretrained model version requires repeating expensive alignment processes. This challenge also applies to domain- or language-specific models, where fine-tuning on specialized data must be redone for every new base model release. In this paper, we explore the transfer of fine-tuning updates between model versions. Specifically, we derive the diff vector from one source model version, which represents the weight changes from fine-tuning, and apply it to the base model of a different target version. Through empirical evaluations on various open-weight model versions, we show that transferring diff vectors can significantly improve the target base model, often achieving performance comparable to its fine-tuned counterpart. For example, reusing the fine-tuning updates from Llama 3.0 8B leads to an absolute accuracy improvement of 10.7% on GPQA over the base Llama 3.1 8B without additional training, surpassing Llama 3.1 8B Instruct. In a multilingual model development setting, we show that this approach can significantly increase performance on target-language tasks without retraining, achieving an absolute improvement of 4.7% and 15.5% on Global MMLU for Malagasy and Turkish, respectively, compared to Llama 3.1 8B Instruct. Our controlled experiments reveal that fine-tuning transfer is most effective when the source and target models are linearly connected in the parameter space. Additionally, we demonstrate that fine-tuning transfer offers a stronger and more computationally efficient starting point for further fine-tuning. Finally, we propose an iterative recycling-then-finetuning approach for continuous model development, which improves both efficiency and effectiveness. Our findings suggest that fine-tuning transfer is a viable strategy to reduce training costs while maintaining model performance.
Self-attention has revolutionized classical machine learning, yet existing quantum self-attention models underutilize quantum states' potential due to oversimplified or incomplete mechanisms. To address this limitation, we introduce the Quantum Complex-Valued Self-Attention Model (QCSAM), the first framework to leverage complex-valued similarities, which captures amplitude and phase relationships between quantum states more comprehensively. To achieve this, QCSAM extends the Linear Combination of Unitaries (LCUs) into the Complex LCUs (CLCUs) framework, enabling precise complex-valued weighting of quantum states and supporting quantum multi-head attention. Experiments on MNIST and Fashion-MNIST show that QCSAM outperforms recent quantum self-attention models, including QKSAN, QSAN, and GQHAN. With only 4 qubits, QCSAM achieves 100% and 99.2% test accuracies on MNIST and Fashion-MNIST, respectively. Furthermore, we evaluate scalability across 3-8 qubits and 2-4 class tasks, while ablation studies validate the advantages of complex-valued attention weights over real-valued alternatives. This work advances quantum machine learning by enhancing the expressiveness and precision of quantum self-attention in a way that aligns with the inherent complexity of quantum mechanics.
Offline optimization is a fundamental challenge in science and engineering, where the goal is to optimize black-box functions using only offline datasets. This setting is particularly relevant when querying the objective function is prohibitively expensive or infeasible, with applications spanning protein engineering, material discovery, neural architecture search, and beyond. The main difficulty lies in accurately estimating the objective landscape beyond the available data, where extrapolations are fraught with significant epistemic uncertainty. This uncertainty can lead to objective hacking(reward hacking), exploiting model inaccuracies in unseen regions, or other spurious optimizations that yield misleadingly high performance estimates outside the training distribution. Recent advances in model-based optimization(MBO) have harnessed the generalization capabilities of deep neural networks to develop offline-specific surrogate and generative models. Trained with carefully designed strategies, these models are more robust against out-of-distribution issues, facilitating the discovery of improved designs. Despite its growing impact in accelerating scientific discovery, the field lacks a comprehensive review. To bridge this gap, we present the first thorough review of offline MBO. We begin by formalizing the problem for both single-objective and multi-objective settings and by reviewing recent benchmarks and evaluation metrics. We then categorize existing approaches into two key areas: surrogate modeling, which emphasizes accurate function approximation in out-of-distribution regions, and generative modeling, which explores high-dimensional design spaces to identify high-performing designs. Finally, we examine the key challenges and propose promising directions for advancement in this rapidly evolving field including safe control of superintelligent systems.
In this work, we propose a Model Predictive Control (MPC) formulation incorporating two distinct horizons: a prediction horizon and a constraint horizon. This approach enables a deeper understanding of how constraints influence key system properties such as suboptimality, without compromising recursive feasibility and constraint satisfaction. In this direction, our contributions are twofold. First, we provide a framework to estimate closed-loop optimality as a function of the number of enforced constraints. This is a generalization of existing results by considering partial constraint enforcement over the prediction horizon. Second, when adopting this general framework under the lens of safety-critical applications, our method improves conventional Control Barrier Function (CBF) based approaches. It mitigates myopic behaviour in Quadratic Programming (QP)-CBF schemes, and resolves compatibility issues between Control Lyapunov Function (CLF) and CBF constraints via the prediction horizon used in the optimization. We show the efficacy of the method via numerical simulations for a safety critical application.