Search

Recent advances in text-based large language models (LLMs), particularly in the GPT series and the o1 model, have demonstrated the effectiveness of scaling both training-time and inference-time compute. However, current state-of-the-art TTS systems leveraging LLMs are often multi-stage, requiring separate models (e.g., diffusion models after LLM), complicating the decision of whether to scale a particular model during training or testing. This work makes the following contributions: First, we explore the scaling of train-time and inference-time compute for speech synthesis. Second, we propose a simple framework Llasa for speech synthesis that employs a single-layer vector quantizer (VQ) codec and a single Transformer architecture to fully align with standard LLMs such as Llama. Our experiments reveal that scaling train-time compute for Llasa consistently improves the naturalness of synthesized speech and enables the generation of more complex and accurate prosody patterns. Furthermore, from the perspective of scaling inference-time compute, we employ speech understanding models as verifiers during the search, finding that scaling inference-time compute shifts the sampling modes toward the preferences of specific verifiers, thereby improving emotional expressiveness, timbre consistency, and content accuracy. In addition, we released the checkpoint and training code for our TTS model (1B, 3B, 8B) and codec model publicly available.

The first measurement of the $CP$ asymmetry of the decay rate ($A_{CP}$) and the $CP$ average ($\Sigma A_{\text{FB}}$) and $CP$ asymmetry ($\Delta A_{\text{FB}}$) of the forward-backward asymmetry in the muon system of $\mathit{\Lambda}^+_c\to p\mu^+\mu^-$ decays is reported. The measurement is performed using a data sample of proton-proton collisions, recorded by the LHCb experiment from 2016 to 2018 at a center-of-mass energy of 13$\text{ TeV}$, which corresponds to an integrated luminosity of 5.4$\text{ fb}^{-1}$. The asymmetries are measured in two regions of dimuon mass near the $\phi$-meson mass peak. The dimuon-mass integrated results are \begin{align*} A_{CP} &= (-1.1 \pm 4.0 \pm 0.5)\%,\\ \Sigma A_{\text{FB}} &= (\phantom{-}3.9 \pm 4.0 \pm 0.6)\%,\\ \Delta A_{\text{FB}} &= (\phantom{-}3.1 \pm 4.0 \pm 0.4)\%, \end{align*} where the first uncertainty is statistical and the second systematic. The results are consistent with the conservation of $CP$ symmetry and the Standard Model expectations., Comment: All figures and tables, along with machine-readable versions and any supplementary material and additional information, are available at https://lbfence.cern.ch/alcm/public/analysis/full-details/3473/ (LHCb public pages)

Let $T_1,\ldots, T_m$ be a family of $d\times d$ invertible real matrices with $\|T_i\|<1/2$ for $1\leq i\leq m$. For ${\bf a}=(a_1,\ldots, a_m)\in {\Bbb R}^{md}$, let $\pi^{\bf a}\colon \Sigma=\{1,\ldots, m\}^{\Bbb N}\to {\Bbb R}^d$ denote the coding map associated with the affine IFS $\{T_ix+a_i\}_{i=1}^m$, and let $K^{\bf a}$ denote the attractor of this IFS. Let $W$ be a linear subspace of ${\Bbb R}^d$ and $P_W$ the orthogonal projection onto $W$. We show that for $\mathcal L^{md}$-a.e.~${\bf a}\in {\Bbb R}^{md}$, the Hausdorff and box-counting dimensions of $P_W(K^{\bf a})$ coincide and are determined by the zero point of a certain pressure function associated with $T_1,\ldots, T_m$ and $W$. Moreover, for every ergodic $\sigma$-invariant measure $\mu$ on $\Sigma$ and for $\mathcal L^{md}$-a.e.~${\bf a}\in {\Bbb R}^{md}$, the local dimensions of $(P_W\pi^{\bf a})_*\mu$ exist almost everywhere, here $(P_W\pi^{\bf a})_*\mu$ stands for the push-forward of $\mu$ by $P_W\pi^{\bf a}$. However, $(P_W\pi^{\bf a})_*\mu$ may not be exact dimensional for $\mathcal L^{md}$-a.e.~${\bf a}\in {\Bbb R}^{md}$. Nevertheless, when $\mu$ is a Bernoulli product measure, or more generally, a supermultiplicative ergodic $\sigma$-invariant measure, $(P_W\pi^{\bf a})_*\mu$ is exact dimensional for $\mathcal L^{md}$-a.e.~${\bf a}\in {\Bbb R}^{md}$., Comment: 64 pages

Remote Photoplethysmography (rPPG) is a promising technique to monitor physiological signals such as heart rate from facial videos. However, the labeled facial videos in this research are challenging to collect. Current rPPG research is mainly based on several small public datasets collected in simple environments, which limits the generalization and scale of the AI models. Semi-supervised methods that leverage a small amount of labeled data and abundant unlabeled data can fill this gap for rPPG learning. In this study, a novel semi-supervised learning method named Semi-rPPG that combines curriculum pseudo-labeling and consistency regularization is proposed to extract intrinsic physiological features from unlabelled data without impairing the model from noises. Specifically, a curriculum pseudo-labeling strategy with signal-to-noise ratio (SNR) criteria is proposed to annotate the unlabelled data while adaptively filtering out the low-quality unlabelled data. Besides, a novel consistency regularization term for quasi-periodic signals is proposed through weak and strong augmented clips. To benefit the research on semi-supervised rPPG measurement, we establish a novel semi-supervised benchmark for rPPG learning through intra-dataset and cross-dataset evaluation on four public datasets. The proposed Semi-rPPG method achieves the best results compared with three classical semi-supervised methods under different protocols. Ablation studies are conducted to prove the effectiveness of the proposed methods., Comment: Accepted by IEEE Transactions on Instrumentation and Measurement (TIM)

Large language models have revolutionized natural language processing but face significant challenges of high storage and runtime costs, due to the transformer architecture's reliance on self-attention, particularly the large Key-Value (KV) cache for long-sequence inference. Recent efforts to reduce KV cache size by pruning less critical entries based on attention weights remain empirical and lack formal grounding. This paper presents a formal study on identifying critical KV cache entries by analyzing attention output perturbation. Our analysis reveals that, beyond attention weights, the value states within KV entries and pretrained parameter matrices are also crucial. Based on this, we propose a perturbation-constrained selection algorithm that optimizes the worst-case output perturbation to identify critical entries. Evaluations on the Needle-in-a-Haystack test and Longbench benchmark show our algorithm enhances state-of-the-art cache eviction methods. Further empirical analysis confirms that our algorithm achieves lower output perturbations in over 92% attention heads in Llama model, thereby providing a significant improvement over existing methods.

Universal knowledge representation is a central problem for multivariate time series(MTS) foundation models and yet remains open. This paper investigates this problem from the first principle and it makes four folds of contributions. First, a new empirical finding is revealed: time series with different time granularities (or corresponding frequency resolutions) exhibit distinct joint distributions in the frequency domain. This implies a crucial aspect of learning universal knowledge, one that has been overlooked by previous studies. Second, a novel Fourier knowledge attention mechanism is proposed to enable learning time granularity-aware representations from both the temporal and frequency domains. Third, an autoregressive blank infilling pre-training framework is incorporated to time series analysis for the first time, leading to a generative tasks agnostic pre-training strategy. To this end, we develop the General Time-series Model (GTM), a unified MTS foundation model that addresses the limitation of contemporary time series models, which often require token, pre-training, or model-level customizations for downstream tasks adaption. Fourth, extensive experiments show that GTM outperforms state-of-the-art (SOTA) methods across all generative tasks, including long-term forecasting, anomaly detection, and imputation.

Integrating human feedback to align text-to-speech (TTS) system outputs with human preferences has proven to be an effective approach for enhancing the robustness of language model-based TTS systems. Current approaches primarily focus on using preference data annotated at the utterance level. However, frequent issues that affect the listening experience often only arise in specific segments of audio samples, while other segments are well-generated. In this study, we propose a fine-grained preference optimization approach (FPO) to enhance the robustness of TTS systems. FPO focuses on addressing localized issues in generated samples rather than uniformly optimizing the entire utterance. Specifically, we first analyze the types of issues in generated samples, categorize them into two groups, and propose a selective training loss strategy to optimize preferences based on fine-grained labels for each issue type. Experimental results show that FPO enhances the robustness of zero-shot TTS systems by effectively addressing local issues, significantly reducing the bad case ratio, and improving intelligibility. Furthermore, FPO exhibits superior data efficiency compared with baseline systems, achieving similar performance with fewer training samples., Comment: WIP

Semantic information refers to the meaning conveyed through words, phrases, and contextual relationships within a given linguistic structure. Humans can leverage semantic information, such as familiar linguistic patterns and contextual cues, to reconstruct incomplete or masked speech signals in noisy environments. However, existing speech enhancement (SE) approaches often overlook the rich semantic information embedded in speech, which is crucial for improving intelligibility, speaker consistency, and overall quality of enhanced speech signals. To enrich the SE model with semantic information, we employ language models as an efficient semantic learner and propose a comprehensive framework tailored for language model-based speech enhancement, called \textit{GenSE}. Specifically, we approach SE as a conditional language modeling task rather than a continuous signal regression problem defined in existing works. This is achieved by tokenizing speech signals into semantic tokens using a pre-trained self-supervised model and into acoustic tokens using a custom-designed single-quantizer neural codec model. To improve the stability of language model predictions, we propose a hierarchical modeling method that decouples the generation of clean semantic tokens and clean acoustic tokens into two distinct stages. Moreover, we introduce a token chain prompting mechanism during the acoustic token generation stage to ensure timbre consistency throughout the speech enhancement process. Experimental results on benchmark datasets demonstrate that our proposed approach outperforms state-of-the-art SE systems in terms of speech quality and generalization capability., Comment: Accepted by ICLR 2025

Reward design for reinforcement learning and optimal control agents is challenging. Preference-based alignment addresses this by enabling agents to learn rewards from ranked trajectory pairs provided by humans. However, existing methods often struggle from poor robustness to unknown false human preferences. In this work, we propose a robust and efficient reward alignment method based on a novel and geometrically interpretable perspective: hypothesis space batched cutting. Our method iteratively refines the reward hypothesis space through "cuts" based on batches of human preferences. Within each batch, human preferences, queried based on disagreement, are grouped using a voting function to determine the appropriate cut, ensuring a bounded human query complexity. To handle unknown erroneous preferences, we introduce a conservative cutting method within each batch, preventing erroneous human preferences from making overly aggressive cuts to the hypothesis space. This guarantees provable robustness against false preferences. We evaluate our method in a model predictive control setting across diverse tasks, including DM-Control, dexterous in-hand manipulation, and locomotion. The results demonstrate that our framework achieves comparable or superior performance to state-of-the-art methods in error-free settings while significantly outperforming existing method when handling high percentage of erroneous human preferences., Comment: 17 pages, including appendix

Since the introduction of Vision Transformer (ViT), patchification has long been regarded as a de facto image tokenization approach for plain visual architectures. By compressing the spatial size of images, this approach can effectively shorten the token sequence and reduce the computational cost of ViT-like plain architectures. In this work, we aim to thoroughly examine the information loss caused by this patchification-based compressive encoding paradigm and how it affects visual understanding. We conduct extensive patch size scaling experiments and excitedly observe an intriguing scaling law in patchification: the models can consistently benefit from decreased patch sizes and attain improved predictive performance, until it reaches the minimum patch size of 1x1, i.e., pixel tokenization. This conclusion is broadly applicable across different vision tasks, various input scales, and diverse architectures such as ViT and the recent Mamba models. Moreover, as a by-product, we discover that with smaller patches, task-specific decoder heads become less critical for dense prediction. In the experiments, we successfully scale up the visual sequence to an exceptional length of 50,176 tokens, achieving a competitive test accuracy of 84.6% with a base-sized model on the ImageNet-1k benchmark. We hope this study can provide insights and theoretical foundations for future works of building non-compressive vision models. Code is available at https://github.com/wangf3014/Patch_Scaling.

The widespread application of autonomous driving technology has significantly advanced the field of autonomous racing. Model Predictive Contouring Control (MPCC) is a highly effective local trajectory planning method for autonomous racing. However, the traditional MPCC method struggles with racetracks that have significant curvature changes, limiting the performance of the vehicle during autonomous racing. To address this issue, we propose a curvature-integrated MPCC (CiMPCC) local trajectory planning method for autonomous racing. This method optimizes the velocity of the local trajectory based on the curvature of the racetrack centerline. The specific implementation involves mapping the curvature of the racetrack centerline to a reference velocity profile, which is then incorporated into the cost function for optimizing the velocity of the local trajectory. This reference velocity profile is created by normalizing and mapping the curvature of the racetrack centerline, thereby ensuring efficient and performance-oriented local trajectory planning in racetracks with significant curvature. The proposed CiMPCC method has been experimented on a self-built 1:10 scale F1TENTH racing vehicle deployed with ROS platform. The experimental results demonstrate that the proposed method achieves outstanding results on a challenging racetrack with sharp curvature, improving the overall lap time by 11.4%-12.5% compared to other autonomous racing trajectory planning methods. Our code is available at https://github.com/zhouhengli/CiMPCC.

Recent advances in large models have significantly advanced image-to-3D reconstruction. However, the generated models are often fused into a single piece, limiting their applicability in downstream tasks. This paper focuses on 3D garment generation, a key area for applications like virtual try-on with dynamic garment animations, which require garments to be separable and simulation-ready. We introduce Dress-1-to-3, a novel pipeline that reconstructs physics-plausible, simulation-ready separated garments with sewing patterns and humans from an in-the-wild image. Starting with the image, our approach combines a pre-trained image-to-sewing pattern generation model for creating coarse sewing patterns with a pre-trained multi-view diffusion model to produce multi-view images. The sewing pattern is further refined using a differentiable garment simulator based on the generated multi-view images. Versatile experiments demonstrate that our optimization approach substantially enhances the geometric alignment of the reconstructed 3D garments and humans with the input image. Furthermore, by integrating a texture generation module and a human motion generation module, we produce customized physics-plausible and realistic dynamic garment demonstrations. Project page: https://dress-1-to-3.github.io/, Comment: Project page: https://dress-1-to-3.github.io/

We propose a generative model of temporally-evolving hypergraphs in which hyperedges form via noisy copying of previous hyperedges. Our proposed model reproduces several stylized facts from many empirical hypergraphs, is learnable from data, and defines a likelihood over a complete hypergraph rather than ego-based or other sub-hypergraphs. Analyzing our model, we derive descriptions of node degree, edge size, and edge intersection size distributions in terms of the model parameters. We also show several features of empirical hypergraphs which are and are not successfully captured by our model. We provide a scalable stochastic expectation maximization algorithm with which we can fit our model to hypergraph data sets with millions of nodes and edges. Finally, we assess our model on a hypergraph link prediction task, finding that an instantiation of our model with just 11 parameters can achieve competitive predictive performance with large neural networks.

4D medical image interpolation is essential for improving temporal resolution and diagnostic precision in clinical applications. Previous works ignore the problem of distribution shifts, resulting in poor generalization under different distribution. A natural solution would be to adapt the model to a new test distribution, but this cannot be done if the test input comes without a ground truth label. In this paper, we propose a novel test time training framework which uses self-supervision to adapt the model to a new distribution without requiring any labels. Indeed, before performing frame interpolation on each test video, the model is trained on the same instance using a self-supervised task, such as rotation prediction or image reconstruction. We conduct experiments on two publicly available 4D medical image interpolation datasets, Cardiac and 4D-Lung. The experimental results show that the proposed method achieves significant performance across various evaluation metrics on both datasets. It achieves higher peak signal-to-noise ratio values, 33.73dB on Cardiac and 34.02dB on 4D-Lung. Our method not only advances 4D medical image interpolation but also provides a template for domain adaptation in other fields such as image segmentation and image registration.

The contextual multi-armed bandit (MAB) is a widely used framework for problems requiring sequential decision-making under uncertainty, such as recommendation systems. In applications involving a large number of users, the performance of contextual MAB can be significantly improved by facilitating collaboration among multiple users. This has been achieved by the clustering of bandits (CB) methods, which adaptively group the users into different clusters and achieve collaboration by allowing the users in the same cluster to share data. However, classical CB algorithms typically rely on numerical reward feedback, which may not be practical in certain real-world applications. For instance, in recommendation systems, it is more realistic and reliable to solicit preference feedback between pairs of recommended items rather than absolute rewards. To address this limitation, we introduce the first "clustering of dueling bandit algorithms" to enable collaborative decision-making based on preference feedback. We propose two novel algorithms: (1) Clustering of Linear Dueling Bandits (COLDB) which models the user reward functions as linear functions of the context vectors, and (2) Clustering of Neural Dueling Bandits (CONDB) which uses a neural network to model complex, non-linear user reward functions. Both algorithms are supported by rigorous theoretical analyses, demonstrating that user collaboration leads to improved regret bounds. Extensive empirical evaluations on synthetic and real-world datasets further validate the effectiveness of our methods, establishing their potential in real-world applications involving multiple users with preference-based feedback., Comment: Preprint

Recent advancements have highlighted the Mamba framework, a state-space model known for its efficiency in capturing long-range dependencies with linear computational complexity. While Mamba has shown competitive performance in medical image segmentation, it encounters difficulties in modeling local features due to the sporadic nature of traditional location-based scanning methods and the complex, ambiguous boundaries often present in medical images. To overcome these challenges, we propose Uncertainty-Driven Mamba (UD-Mamba), which redefines the pixel-order scanning process by incorporating channel uncertainty into the scanning mechanism. UD-Mamba introduces two key scanning techniques: 1) sequential scanning, which prioritizes regions with high uncertainty by scanning in a row-by-row fashion, and 2) skip scanning, which processes columns vertically, moving from high-to-low or low-to-high uncertainty at fixed intervals. Sequential scanning efficiently clusters high-uncertainty regions, such as boundaries and foreground objects, to improve segmentation precision, while skip scanning enhances the interaction between background and foreground regions, allowing for timely integration of background information to support more accurate foreground inference. Recognizing the advantages of scanning from certain to uncertain areas, we introduce four learnable parameters to balance the importance of features extracted from different scanning methods. Additionally, a cosine consistency loss is employed to mitigate the drawbacks of transitioning between uncertain and certain regions during the scanning process. Our method demonstrates robust segmentation performance, validated across three distinct medical imaging datasets involving pathology, dermatological lesions, and cardiac tasks., Comment: 19 pages

Segment Anything Model 2 (SAM 2), a prompt-driven foundation model extending SAM to both image and video domains, has shown superior zero-shot performance compared to its predecessor. Building on SAM's success in medical image segmentation, SAM 2 presents significant potential for further advancement. However, similar to SAM, SAM 2 is limited by its output of binary masks, inability to infer semantic labels, and dependence on precise prompts for the target object area. Additionally, direct application of SAM and SAM 2 to medical image segmentation tasks yields suboptimal results. In this paper, we explore the upper performance limit of SAM 2 using custom fine-tuning adapters, achieving a Dice Similarity Coefficient (DSC) of 92.30% on the BTCV dataset, surpassing the state-of-the-art nnUNet by 12%. Following this, we address the prompt dependency by investigating various prompt generators. We introduce a UNet to autonomously generate predicted masks and bounding boxes, which serve as input to SAM 2. Subsequent dual-stage refinements by SAM 2 further enhance performance. Extensive experiments show that our method achieves state-of-the-art results on the AMOS2022 dataset, with a Dice improvement of 2.9% compared to nnUNet, and outperforms nnUNet by 6.4% on the BTCV dataset.

ICESat-2 (IS2) by NASA is an Earth-observing satellite that measures high-resolution surface elevation. The IS2's ATL07 and ATL10 sea ice elevation and freeboard products of 10m-200m segments which aggregated 150 signal photons from the raw ATL03 (geolocated photon) data. These aggregated products can potentially overestimate local sea surface height, thus underestimating the calculations of freeboard (sea ice height above sea surface). To achieve a higher resolution of sea surface height and freeboard information, in this work we utilize a 2m window to resample the ATL03 data. Then, we classify these 2m segments into thick sea ice, thin ice, and open water using deep learning methods (Long short-term memory and Multi-layer perceptron models). To obtain labeled training data for our deep learning models, we use segmented Sentinel-2 (S2) multi-spectral imagery overlapping with IS2 tracks in space and time to auto-label IS2 data, followed by some manual corrections in the regions of transition between different ice/water types or cloudy regions. We employ a parallel workflow for this auto-labeling using PySpark to scale, and we achieve 9-fold data loading and 16.25-fold map-reduce speedup. To train our models, we employ a Horovod-based distributed deep-learning workflow on a DGX A100 8 GPU cluster, achieving a 7.25-fold speedup. Next, we calculate the local sea surface heights based on the open water segments. Finally, we scale the freeboard calculation using the derived local sea level and achieve 8.54-fold data loading and 15.7-fold map-reduce speedup. Compared with the ATL07 (local sea level) and ATL10 (freeboard) data products, our results show higher resolutions and accuracy (96.56%).

Controllable video generation remains a significant challenge, despite recent advances in generating high-quality and consistent videos. Most existing methods for controlling video generation treat the video as a whole, neglecting intricate fine-grained spatiotemporal relationships, which limits both control precision and efficiency. In this paper, we propose Controllable Video Generative Adversarial Networks (CoVoGAN) to disentangle the video concepts, thus facilitating efficient and independent control over individual concepts. Specifically, following the minimal change principle, we first disentangle static and dynamic latent variables. We then leverage the sufficient change property to achieve component-wise identifiability of dynamic latent variables, enabling independent control over motion and identity. To establish the theoretical foundation, we provide a rigorous analysis demonstrating the identifiability of our approach. Building on these theoretical insights, we design a Temporal Transition Module to disentangle latent dynamics. To enforce the minimal change principle and sufficient change property, we minimize the dimensionality of latent dynamic variables and impose temporal conditional independence. To validate our approach, we integrate this module as a plug-in for GANs. Extensive qualitative and quantitative experiments on various video generation benchmarks demonstrate that our method significantly improves generation quality and controllability across diverse real-world scenarios.

Numerical simulation of multi-component flow systems characterized by the simultaneous presence of pressure-velocity coupling and pressure-density coupling dominated regions remains a significant challenge in computational fluid dynamics. Thus, this work presents a novel approach that combines the Godunov-type scheme for high-speed flows with the projection solution procedure for incompressible flows to address this challenge. The proposed hybrid approach begins by splitting the inviscid flux into the advection part and the pressure part. The solution variables are first updated to their intermediate states by solving the advection part with the all-speed AUSM (Advection Upwind Splitting Method) Riemann solver. The advection flux in AUSM is modified to eliminate the pressure flux term that deteriorates the accuracy at the low Mach region. To prevent the advection flux from causing spurious velocities when surface tension is present, the pressure-velocity coupling term is modified to ensure it vanishes at material interfaces. Then, we derive the pressure Helmholtz equation to solve the final pressure and update the intermediate states to the solution variables at the next time step. The proposed hybrid approach retains the upwind property of the AUSM scheme for high Mach numbers while recovering central schemes and the standard projection solution for low Mach limits. To accurately resolve the complex flow structures including shock waves and material interfaces without numerical oscillations, a newly proposed homogenous ROUND (Reconstruction Operator on Unified Normalised-variable Diagram) reconstruction strategy is employed in this work. By simulating high-speed compressible multiphase flows and incompressible multiphase flows, this study demonstrates the ability of the proposed method to accurately handle flow regimes across all Mach numbers.

Survival analysis, a vital tool for predicting the time to event, has been used in many domains such as healthcare, criminal justice, and finance. Like classification tasks, survival analysis can exhibit bias against disadvantaged groups, often due to biases inherent in data or algorithms. Several studies in both the IS and CS communities have attempted to address fairness in survival analysis. However, existing methods often overlook the importance of prediction fairness at pre-defined evaluation time points, which is crucial in real-world applications where decision making often hinges on specific time frames. To address this critical research gap, we introduce a new fairness concept: equalized odds (EO) in survival analysis, which emphasizes prediction fairness at pre-defined time points. To achieve the EO fairness in survival analysis, we propose a Conditional Mutual Information Augmentation (CMIA) approach, which features a novel fairness regularization term based on conditional mutual information and an innovative censored data augmentation technique. Our CMIA approach can effectively balance prediction accuracy and fairness, and it is applicable to various survival models. We evaluate the CMIA approach against several state-of-the-art methods within three different application domains, and the results demonstrate that CMIA consistently reduces prediction disparity while maintaining good accuracy and significantly outperforms the other competing methods across multiple datasets and survival models (e.g., linear COX, deep AFT).

Values are core drivers of individual and collective perception, cognition, and behavior. Value systems, such as Schwartz's Theory of Basic Human Values, delineate the hierarchy and interplay among these values, enabling cross-disciplinary investigations into decision-making and societal dynamics. Recently, the rise of Large Language Models (LLMs) has raised concerns regarding their elusive intrinsic values. Despite growing efforts in evaluating, understanding, and aligning LLM values, a psychologically grounded LLM value system remains underexplored. This study addresses the gap by introducing the Generative Psycho-Lexical Approach (GPLA), a scalable, adaptable, and theoretically informed method for constructing value systems. Leveraging GPLA, we propose a psychologically grounded five-factor value system tailored for LLMs. For systematic validation, we present three benchmarking tasks that integrate psychological principles with cutting-edge AI priorities. Our results reveal that the proposed value system meets standard psychological criteria, better captures LLM values, improves LLM safety prediction, and enhances LLM alignment, when compared to the canonical Schwartz's values.

Teaching large language models (LLMs) to critique and refine their outputs is crucial for building systems that can iteratively improve, yet it is fundamentally limited by the ability to provide accurate judgments and actionable suggestions. In this work, we study LLM critics for code generation and propose $\texttt{CTRL}$, a framework for $\texttt{C}$ritic $\texttt{T}$raining via $\texttt{R}$einforcement $\texttt{L}$earning, which trains a critic model to generate feedback that maximizes correction performance for a fixed generator model without human supervision. Our results demonstrate that critics trained with $\texttt{CTRL}$ significantly enhance pass rates and mitigate compounding errors across both base and stronger generator models. Furthermore, we show that these critic models act as accurate generative reward models and enable test-time scaling through iterative critique-revision, achieving up to 106.1% relative improvements across challenging code generation benchmarks.

This paper presents a novel approach to Visual Inertial Odometry (VIO), focusing on the initialization and feature matching modules. Existing methods for initialization often suffer from either poor stability in visual Structure from Motion (SfM) or fragility in solving a huge number of parameters simultaneously. To address these challenges, we propose a new pipeline for visual inertial initialization that robustly handles various complex scenarios. By tightly coupling gyroscope measurements, we enhance the robustness and accuracy of visual SfM. Our method demonstrates stable performance even with only four image frames, yielding competitive results. In terms of feature matching, we introduce a hybrid method that combines optical flow and descriptor-based matching. By leveraging the robustness of continuous optical flow tracking and the accuracy of descriptor matching, our approach achieves efficient, accurate, and robust tracking results. Through evaluation on multiple benchmarks, our method demonstrates state-of-the-art performance in terms of accuracy and success rate. Additionally, a video demonstration on mobile devices showcases the practical applicability of our approach in the field of Augmented Reality/Virtual Reality (AR/VR).

Molecular odor prediction has great potential across diverse fields such as chemistry, pharmaceuticals, and environmental science, enabling the rapid design of new materials and enhancing environmental monitoring. However, current methods face two main challenges: First, existing models struggle with non-smooth objective functions and the complexity of mixed feature dimensions; Second, datasets suffer from severe label imbalance, which hampers model training, particularly in learning minority class labels. To address these issues, we introduce a novel feature mapping method and a molecular ensemble optimization loss function. By incorporating feature importance learning and frequency modulation, our model adaptively adjusts the contribution of each feature, efficiently capturing the intricate relationship between molecular structures and odor descriptors. Our feature mapping preserves feature independence while enhancing the model's efficiency in utilizing molecular features through frequency modulation. Furthermore, the proposed loss function dynamically adjusts label weights, improves structural consistency, and strengthens label correlations, effectively addressing data imbalance and label co-occurrence challenges. Experimental results show that our method significantly can improves the accuracy of molecular odor prediction across various deep learning models, demonstrating its promising potential in molecular structure representation and chemoinformatics.

Multi-label learning (MLL) has gained attention for its ability to represent real-world data. Label Distribution Learning (LDL), an extension of MLL to learning from label distributions, faces challenges in collecting accurate label distributions. To address the issue of biased annotations, based on the low-rank assumption, existing works recover true distributions from biased observations by exploring the label correlations. However, recent evidence shows that the label distribution tends to be full-rank, and naive apply of low-rank approximation on biased observation leads to inaccurate recovery and performance degradation. In this paper, we address the LDL with biased annotations problem from a novel perspective, where we first degenerate the soft label distribution into a hard multi-hot label and then recover the true label information for each instance. This idea stems from an insight that assigning hard multi-hot labels is often easier than assigning a soft label distribution, and it shows stronger immunity to noise disturbances, leading to smaller label bias. Moreover, assuming that the multi-label space for predicting label distributions is low-rank offers a more reasonable approach to capturing label correlations. Theoretical analysis and experiments confirm the effectiveness and robustness of our method on real-world datasets.

Chinese literary classics hold significant cultural and educational value, offering deep insights into morality, history, and human nature. These works often include classical Chinese and complex narratives, making them difficult for children to read. To bridge this gap, we introduce a child-friendly literary adaptation (CLA) task to adapt the Chinese literary classic into engaging and accessible text for children. However, recent large language models (LLMs) overlook children's reading preferences (\ie, vivid character portrayals, concise narrative structures, and appropriate readability), which poses challenges in CLA. In this paper, we propose a method called InstructChild, which augments the LLM with these preferences for adaptation. Specifically, we first obtain the characters' personalities and narrative structure as additional information for fine-grained instruction tuning. Then, we devise a readability metric as the reward to align the LLM with the children's reading level. Finally, a lookahead decoding strategy is applied to improve the readability of the generated text during inference. To support the evaluation of CLA task, we construct the Classic4Children dataset, which comprises both the original and child-friendly versions of the Four Great Classical Novels of Chinese literature. Experimental results show that our InstructChild significantly improves automatic and human evaluation performance., Comment: Accepted at NAACL 2025 Findings

Diffusion models suffer from the huge consumption of time and resources to train. For example, diffusion models need hundreds of GPUs to train for several weeks for a high-resolution generative task to meet the requirements of an extremely large number of iterations and a large batch size. Training diffusion models become a millionaire's game. With limited resources that only fit a small batch size, training a diffusion model always fails. In this paper, we investigate the key reasons behind the difficulties of training diffusion models with limited resources. Through numerous experiments and demonstrations, we identified a major factor: the significant variation in the training losses across different timesteps, which can easily disrupt the progress made in previous iterations. Moreover, different prediction types of $x_0$ exhibit varying effectiveness depending on the task and timestep. We hypothesize that using a mixed-prediction approach to identify the most accurate $x_0$ prediction type could potentially serve as a breakthrough in addressing this issue. In this paper, we outline several challenges and insights, with the hope of inspiring further research aimed at tackling the limitations of training diffusion models with constrained resources, particularly for high-resolution tasks.

Optimal Transport (OT) theory seeks to determine the map $T:X \to Y$ that transports a source measure $P$ to a target measure $Q$, minimizing the cost $c(\mathbf{x}, T(\mathbf{x}))$ between $\mathbf{x}$ and its image $T(\mathbf{x})$. Building upon the Input Convex Neural Network OT solver and incorporating the concept of displacement-sparse maps, we introduce a sparsity penalty into the minimax Wasserstein formulation, promote sparsity in displacement vectors $\Delta(\mathbf{x}) := T(\mathbf{x}) - \mathbf{x}$, and enhance the interpretability of the resulting map. However, increasing sparsity often reduces feasibility, causing $T_{\#}(P)$ to deviate more significantly from the target measure. In low-dimensional settings, we propose a heuristic framework to balance the trade-off between sparsity and feasibility by dynamically adjusting the sparsity intensity parameter during training. For high-dimensional settings, we directly constrain the dimensionality of displacement vectors by enforcing $\dim(\Delta(\mathbf{x})) \leq l$, where $l < d$ for $X \subseteq \mathbb{R}^d$. Among maps satisfying this constraint, we aim to identify the most feasible one. This goal can be effectively achieved by adapting our low-dimensional heuristic framework without resorting to dimensionality reduction. We validate our method on both synthesized sc-RNA and real 4i cell perturbation datasets, demonstrating improvements over existing methods., Comment: 18 pages, 6 Figures

The quenching of star formation activity represents a critical phase for a non-negligible fraction of the observed galaxy population at all cosmic epochs, marking a transition from an epoch of intense mass growth to an extended period of passive evolution. Over the past years, we have collected a detailed characterization of how the fraction of quenched galaxies evolves as a function of cosmic time (it grows at later cosmic epochs), and correlates with the physical properties of galaxies (more massive galaxies, that also tend to be bulge-dominated, are more likely to be quenched) and with their environment (denser environments host larger fractions of quiescent galaxies). Different physical processes can lead to a suppression of star formation. These include internal processes (processes that do not depend on the environment in which galaxies live, e.g. stellar and AGN feedback mechanisms) and processes that instead depend on the environment (e.g. mergers, gas stripping processes). In this chapter, we summarize the observational and theoretical status of the field, highlighting the most recent results and the questions that are yet to be answered. We also summarize our view on the expected developments in the next few years., Comment: This is a pre-print of a chapter for the Encyclopedia of Astrophysics (edited by I. Mandel, section editor S. McGee) to be published by Elsevier as a Reference Module

The integration of vision-language modalities has been a significant focus in multimodal learning, traditionally relying on Vision-Language Pretrained Models. However, with the advent of Large Language Models (LLMs), there has been a notable shift towards incorporating LLMs with vision modalities. Following this, the training paradigms for incorporating vision modalities into LLMs have evolved. Initially, the approach was to integrate the modalities through pretraining the modality integrator, named Single-stage Tuning. It has since branched out into methods focusing on performance enhancement, denoted as Two-stage Tuning, and those prioritizing parameter efficiency, referred to as Direct Adaptation. However, existing surveys primarily address the latest Vision Large Language Models (VLLMs) with Two-stage Tuning, leaving a gap in understanding the evolution of training paradigms and their unique parameter-efficient considerations. This paper categorizes and reviews 34 VLLMs from top conferences, journals, and highly cited Arxiv papers, focusing on parameter efficiency during adaptation from the training paradigm perspective. We first introduce the architecture of LLMs and parameter-efficient learning methods, followed by a discussion on vision encoders and a comprehensive taxonomy of modality integrators. We then review three training paradigms and their efficiency considerations, summarizing benchmarks in the VLLM field. To gain deeper insights into their effectiveness in parameter efficiency, we compare and discuss the experimental results of representative models, among which the experiment of the Direct Adaptation paradigm is replicated. Providing insights into recent developments and practical uses, this survey is a vital guide for researchers and practitioners navigating the efficient integration of vision modalities into LLMs., Comment: 28 pages, 3 figures

Large language models (LLMs) are prone to hallucinations in question-answering (QA) tasks when faced with ambiguous questions. Users often assume that LLMs share their cognitive alignment, a mutual understanding of context, intent, and implicit details, leading them to omit critical information in the queries. However, LLMs generate responses based on assumptions that can misalign with user intent, which may be perceived as hallucinations if they misalign with the user's intent. Therefore, identifying those implicit assumptions is crucial to resolve ambiguities in QA. Prior work, such as AmbigQA, reduces ambiguity in queries via human-annotated clarifications, which is not feasible in real application. Meanwhile, ASQA compiles AmbigQA's short answers into long-form responses but inherits human biases and fails capture explicit logical distinctions that differentiates the answers. We introduce Conditional Ambiguous Question-Answering (CondAmbigQA), a benchmark with 200 ambiguous queries and condition-aware evaluation metrics. Our study pioneers the concept of ``conditions'' in ambiguous QA tasks, where conditions stand for contextual constraints or assumptions that resolve ambiguities. The retrieval-based annotation strategy uses retrieved Wikipedia fragments to identify possible interpretations for a given query as its conditions and annotate the answers through those conditions. Such a strategy minimizes human bias introduced by different knowledge levels among annotators. By fixing retrieval results, CondAmbigQA evaluates how RAG systems leverage conditions to resolve ambiguities. Experiments show that models considering conditions before answering improve performance by $20\%$, with an additional $5\%$ gain when conditions are explicitly provided. These results underscore the value of conditional reasoning in QA, offering researchers tools to rigorously evaluate ambiguity resolution.

Olfactory perception plays a critical role in both human and organismal interactions, yet understanding of its underlying mechanisms and influencing factors remain insufficient. Molecular structures influence odor perception through intricate biochemical interactions, and accurately quantifying structure-odor relationships presents significant challenges. The Quantitative Structure-Odor Relationship (QSOR) task, which involves predicting the associations between molecular structures and their corresponding odors, seeks to address these challenges. To this end, we propose a method for QSOR, utilizing Graph Attention Networks to model molecular structures and capture both local and global features. Unlike conventional QSOR approaches reliant on predefined descriptors, our method leverages diverse molecular feature extraction techniques to automatically learn comprehensive representations. This integration enhances the model's capacity to handle complex molecular information, improves prediction accuracy. Our approach demonstrates clear advantages in QSOR prediction tasks, offering valuable insights into the application of deep learning in cheminformatics.

Large language models (LLMs) process entire input contexts indiscriminately, which is inefficient in cases where the information required to answer a query is localized within the context. We present dynamic context cutoff, a human-inspired method enabling LLMs to self-terminate processing upon acquiring sufficient task-relevant information. Through analysis of model internals, we discover that specific attention heads inherently encode "sufficiency signals" - detectable through lightweight classifiers - that predict when critical information has been processed. This reveals a new efficiency paradigm: models' internal understanding naturally dictates processing needs rather than external compression heuristics. Comprehensive experiments across six QA datasets (up to 40K tokens) with three model families (LLaMA/Qwen/Mistral, 1B0-70B) demonstrate 1.33x average token reduction while improving accuracy by 1.3%. Furthermore, our method demonstrates better performance with the same rate of token reduction compared to other context efficiency methods. Additionally, we observe an emergent scaling phenomenon: while smaller models require require probing for sufficiency detection, larger models exhibit intrinsic self-assessment capabilities through prompting., Comment: Project Website: https://royxie.com/when-to-stop-project/

We consider detecting change points in the correlation structure of streaming large-dimensional data with minimum assumptions posed on the underlying data distribution. Depending on the $\ell_1$ and $\ell_{\infty}$ norms of the squared difference of vectorized pre-change and post-change correlation matrices, detection statistics are constructed for dense and sparse settings, respectively. The proposed detection procedures possess the bless-dimension property, as a novel algorithm for threshold selection is designed based on sign-flip permutation. Theoretical evaluations of the proposed methods are conducted in terms of average run length and expected detection delay. Numerical studies are conducted to examine the finite sample performances of the proposed methods. Our methods are effective because the average detection delays have slopes similar to that of the optimal exact CUSUM test. Moreover, a combined $\ell_1$ and $\ell_{\infty}$ norm approach is proposed and has expected performance for transitions from sparse to dense settings. Our method is applied to forecast El Ni{\~n}o events and achieves state-of-the-art hit rates greater than 0.86, while false alarm rates are 0. This application illustrates the efficiency and effectiveness of our proposed methodology in detecting fundamental changes with minimal delay.

In an Information Retrieval (IR) system, reranking plays a critical role by sorting candidate passages according to their relevance to a specific query. This process demands a nuanced understanding of the variations among passages linked to the query. In this work, we introduce RankFlow, a multi-role reranking workflow that leverages the capabilities of Large Language Models (LLMs) and role specializations to improve reranking performance. RankFlow enlists LLMs to fulfill four distinct roles: the query Rewriter, the pseudo Answerer, the passage Summarizer, and the Reranker. This orchestrated approach enables RankFlow to: (1) accurately interpret queries, (2) draw upon LLMs' extensive pre-existing knowledge, (3) distill passages into concise versions, and (4) assess passages in a comprehensive manner, resulting in notably better reranking results. Our experimental results reveal that RankFlow outperforms existing leading approaches on widely recognized IR benchmarks, such as TREC-DL, BEIR, and NovelEval. Additionally, we investigate the individual contributions of each role in RankFlow. Code is available at https://github.com/jincan333/RankFlow.

Can integrating spectral and curvature signals unlock new potential in graph representation learning? Non-Euclidean geometries, particularly Riemannian manifolds such as hyperbolic (negative curvature) and spherical (positive curvature), offer powerful inductive biases for embedding complex graph structures like scale-free, hierarchical, and cyclic patterns. Meanwhile, spectral filtering excels at processing signal variations across graphs, making it effective in homophilic and heterophilic settings. Leveraging both can significantly enhance the learned representations. To this end, we propose Spectro-Riemannian Graph Neural Networks (CUSP) - the first graph representation learning paradigm that unifies both CUrvature (geometric) and SPectral insights. CUSP is a mixed-curvature spectral GNN that learns spectral filters to optimize node embeddings in products of constant-curvature manifolds (hyperbolic, spherical, and Euclidean). Specifically, CUSP introduces three novel components: (a) Cusp Laplacian, an extension of the traditional graph Laplacian based on Ollivier-Ricci curvature, designed to capture the curvature signals better; (b) Cusp Filtering, which employs multiple Riemannian graph filters to obtain cues from various bands in the eigenspectrum; and (c) Cusp Pooling, a hierarchical attention mechanism combined with a curvature-based positional encoding to assess the relative importance of differently curved substructures in our graph. Empirical evaluation across eight homophilic and heterophilic datasets demonstrates the superiority of CUSP in node classification and link prediction tasks, with a gain of up to 5.3% over state-of-the-art models., Comment: ICLR 2025

Federated learning (FL) has gained widespread attention for its privacy-preserving and collaborative learning capabilities. Due to significant statistical heterogeneity, traditional FL struggles to generalize a shared model across diverse data domains. Personalized federated learning addresses this issue by dividing the model into a globally shared part and a locally private part, with the local model correcting representation biases introduced by the global model. Nevertheless, locally converged parameters more accurately capture domain-specific knowledge, and current methods overlook the potential benefits of these parameters. To address these limitations, we propose PM-MoE architecture. This architecture integrates a mixture of personalized modules and an energy-based personalized modules denoising, enabling each client to select beneficial personalized parameters from other clients. We applied the PM-MoE architecture to nine recent model-split-based personalized federated learning algorithms, achieving performance improvements with minimal additional training. Extensive experiments on six widely adopted datasets and two heterogeneity settings validate the effectiveness of our approach. The source code is available at \url{https://github.com/dannis97500/PM-MOE}.

Segment Anything Model (SAM) has demonstrated impressive zero-shot performance and brought a range of unexplored capabilities to natural image segmentation tasks. However, as a very important branch of image segmentation, the performance of SAM remains uncertain when applied to medical image segmentation due to the significant differences between natural images and medical images. Meanwhile, it is harsh to meet the SAM's requirements of extra prompts provided, such as points or boxes to specify medical regions. In this paper, we propose a novel self-prompt SAM adaptation framework for medical image segmentation, named Self-Prompt-SAM. We design a multi-scale prompt generator combined with the image encoder in SAM to generate auxiliary masks. Then, we use the auxiliary masks to generate bounding boxes as box prompts and use Distance Transform to select the most central points as point prompts. Meanwhile, we design a 3D depth-fused adapter (DfusedAdapter) and inject the DFusedAdapter into each transformer in the image encoder and mask decoder to enable pre-trained 2D SAM models to extract 3D information and adapt to 3D medical images. Extensive experiments demonstrate that our method achieves state-of-the-art performance and outperforms nnUNet by 2.3% on AMOS2022, 1.6% on ACDCand 0.5% on Synapse datasets.

Emotions play a crucial role in human behavior and decision-making, making emotion recognition a key area of interest in human-computer interaction (HCI). This study addresses the challenges of emotion recognition by integrating facial expression analysis with electroencephalogram (EEG) signals, introducing a novel multimodal framework-Milmer. The proposed framework employs a transformer-based fusion approach to effectively integrate visual and physiological modalities. It consists of an EEG preprocessing module, a facial feature extraction and balancing module, and a cross-modal fusion module. To enhance visual feature extraction, we fine-tune a pre-trained Swin Transformer on emotion-related datasets. Additionally, a cross-attention mechanism is introduced to balance token representation across modalities, ensuring effective feature integration. A key innovation of this work is the adoption of a multiple instance learning (MIL) approach, which extracts meaningful information from multiple facial expression images over time, capturing critical temporal dynamics often overlooked in previous studies. Extensive experiments conducted on the DEAP dataset demonstrate the superiority of the proposed framework, achieving a classification accuracy of 96.72% in the four-class emotion recognition task. Ablation studies further validate the contributions of each module, highlighting the significance of advanced feature extraction and fusion strategies in enhancing emotion recognition performance. Our code are available at https://github.com/liangyubuaa/Milmer.

The goal of diffusion generative models is to align the learned distribution with the real data distribution through gradient score matching. However, inherent limitations in training data quality, modeling strategies, and architectural design lead to inevitable gap between generated outputs and real data. To reduce this gap, we propose Weak-to-Strong Diffusion (W2SD), a novel framework that utilizes the estimated difference between existing weak and strong models (i.e., weak-to-strong difference) to approximate the gap between an ideal model and a strong model. By employing a reflective operation that alternates between denoising and inversion with weak-to-strong difference, we theoretically understand that W2SD steers latent variables along sampling trajectories toward regions of the real data distribution. W2SD is highly flexible and broadly applicable, enabling diverse improvements through the strategic selection of weak-to-strong model pairs (e.g., DreamShaper vs. SD1.5, good experts vs. bad experts in MoE). Extensive experiments demonstrate that W2SD significantly improves human preference, aesthetic quality, and prompt adherence, achieving SOTA performance across various modalities (e.g., image, video), architectures (e.g., UNet-based, DiT-based, MoE), and benchmarks. For example, Juggernaut-XL with W2SD can improve with the HPSv2 winning rate up to 90% over the original results. Moreover, the performance gains achieved by W2SD markedly outweigh its additional computational overhead, while the cumulative improvements from different weak-to-strong difference further solidify its practical utility and deployability., Comment: 20 pages, 19 figures, 14 tables

We present a correlation between the stellar metallicities and the mutual inclinations of multi-planet systems hosting short-period small planets (a/Rs<12, Rp<4Re). We analyzed 89 multi-planet systems discovered by Kepler, K2, and TESS, where the innermost planets have periods shorter than 10 days. We found that the mutual inclinations of the innermost two planets are higher and more diverse around metal-rich stars. The mutual inclinations are calculated as the absolute differences between the best-fit inclinations of the innermost two planets from transit modeling, which represent the lower limits of the true mutual inclinations. The mean and variance of the mutual inclination distribution of the metal-rich systems are 3.1+-0.5 and 3.1+-0.4 degrees, while for the metal-poor systems they are 1.3+-0.2 and 1.0+-0.2 degrees. This finding suggests that inner planetary systems around metal-rich stars are dynamically hotter. We summarized the theories that could plausibly explain this correlation, including the influence of giant planets, higher solid densities in protoplanetary disks around metal-rich stars, or secular chaos coupled with an excess of angular momentum deficits. Planet formation and population synthesis models tracking the mutual inclination evolution would be essential to fully understand this correlation., Comment: 14 pages, 5 figures, 1 table. Accepted by ApJL

The next-generation mm/sub-mm/far-IR astronomy will in part be enabled by advanced digital signal processing (DSP) techniques. The Prime-Cam instrument of the Fred Young Submillimeter Telescope (FYST), featuring the largest array of submillimeter detectors to date, utilizes a novel overlap-channel polyphase synthesis filter bank (OC-PSB) for the AC biasing of detectors, implemented on a cutting-edge Xilinx Radio Frequency System on Chip (RFSoC). This design departs from traditional waveform look-up-table(LUT) in memory, allowing real-time, dynamic signal generation, enhancing usable bandwidth and dynamic range, and enabling microwave kinetic inductance detector (MKID) tracking for future readout systems. Results show that the OC-PSB upholds critical performance metrics such as signal-to-noise ratio (SNR) while offering additional benefits such as scalability. This paper will discuss DSP design, RFSoC implementation, and laboratory performance, demonstrating OC-PSB's potential in submillimeter-wave astronomy MKID readout systems., Comment: SPIE Astronomical Telescopes + Instrumentation conference proceedings 2024

$Q$-learning is one of the most fundamental reinforcement learning algorithms. Previously, it is widely believed that $Q$-learning with linear function approximation (i.e., linear $Q$-learning) suffers from possible divergence. This paper instead establishes the first $L^2$ convergence rate of linear $Q$-learning to a bounded set. Notably, we do not make any modification to the original linear $Q$-learning algorithm, do not make any Bellman completeness assumption, and do not make any near-optimality assumption on the behavior policy. All we need is an $\epsilon$-softmax behavior policy with an adaptive temperature. The key to our analysis is the general result of stochastic approximations under Markovian noise with fast-changing transition functions. As a side product, we also use this general result to establish the $L^2$ convergence rate of tabular $Q$-learning with an $\epsilon$-softmax behavior policy, for which we rely on a novel pseudo-contraction property of the weighted Bellman optimality operator.

Domain shifts are critical issues that harm the performance of machine learning. Unsupervised Domain Adaptation (UDA) mitigates this issue but suffers when the domain shifts are steep and drastic. Gradual Domain Adaptation (GDA) alleviates this problem in a mild way by gradually adapting from the source to the target domain using multiple intermediate domains. In this paper, we propose Sliding Window Adversarial Training (SWAT) for Gradual Domain Adaptation. SWAT uses the construction of adversarial streams to connect the feature spaces of the source and target domains. In order to gradually narrow the small gap between adjacent intermediate domains, a sliding window paradigm is designed that moves along the adversarial stream. When the window moves to the end of the stream, i.e., the target domain, the domain shift is drastically reduced. Extensive experiments are conducted on public GDA benchmarks, and the results demonstrate that the proposed SWAT significantly outperforms the state-of-the-art approaches. The implementation is available at: https://anonymous.4open.science/r/SWAT-8677., Comment: submitted to icml 2025

Image Super-Resolution (ISR) has seen significant progress with the introduction of remarkable generative models. However, challenges such as the trade-off issues between fidelity and realism, as well as computational complexity, have also posed limitations on their application. Building upon the tremendous success of autoregressive models in the language domain, we propose \textbf{VARSR}, a novel visual autoregressive modeling for ISR framework with the form of next-scale prediction. To effectively integrate and preserve semantic information in low-resolution images, we propose using prefix tokens to incorporate the condition. Scale-aligned Rotary Positional Encodings are introduced to capture spatial structures and the diffusion refiner is utilized for modeling quantization residual loss to achieve pixel-level fidelity. Image-based Classifier-free Guidance is proposed to guide the generation of more realistic images. Furthermore, we collect large-scale data and design a training process to obtain robust generative priors. Quantitative and qualitative results show that VARSR is capable of generating high-fidelity and high-realism images with more efficiency than diffusion-based methods. Our codes will be released at https://github.com/qyp2000/VARSR., Comment: 20 pages; 17 figures

Recent open-vocabulary detectors achieve promising performance with abundant region-level annotated data. In this work, we show that an open-vocabulary detector co-training with a large language model by generating image-level detailed captions for each image can further improve performance. To achieve the goal, we first collect a dataset, GroundingCap-1M, wherein each image is accompanied by associated grounding labels and an image-level detailed caption. With this dataset, we finetune an open-vocabulary detector with training objectives including a standard grounding loss and a caption generation loss. We take advantage of a large language model to generate both region-level short captions for each region of interest and image-level long captions for the whole image. Under the supervision of the large language model, the resulting detector, LLMDet, outperforms the baseline by a clear margin, enjoying superior open-vocabulary ability. Further, we show that the improved LLMDet can in turn build a stronger large multi-modal model, achieving mutual benefits. The code, model, and dataset is available at https://github.com/iSEE-Laboratory/LLMDet.

In this paper, we consider the problem of long-term point tracking, which requires consistent identification of points across multiple frames in a video, despite changes in appearance, lighting, perspective, and occlusions. We target online tracking on a frame-by-frame basis, making it suitable for real-world, streaming scenarios. Specifically, we introduce Track-On, a simple transformer-based model designed for online long-term point tracking. Unlike prior methods that depend on full temporal modeling, our model processes video frames causally without access to future frames, leveraging two memory modules -- spatial memory and context memory -- to capture temporal information and maintain reliable point tracking over long time horizons. At inference time, it employs patch classification and refinement to identify correspondences and track points with high accuracy. Through extensive experiments, we demonstrate that Track-On sets a new state-of-the-art for online models and delivers superior or competitive results compared to offline approaches on seven datasets, including the TAP-Vid benchmark. Our method offers a robust and scalable solution for real-time tracking in diverse applications. Project page: https://kuis-ai.github.io/track_on, Comment: ICLR 2025

This paper presents SANA-1.5, a linear Diffusion Transformer for efficient scaling in text-to-image generation. Building upon SANA-1.0, we introduce three key innovations: (1) Efficient Training Scaling: A depth-growth paradigm that enables scaling from 1.6B to 4.8B parameters with significantly reduced computational resources, combined with a memory-efficient 8-bit optimizer. (2) Model Depth Pruning: A block importance analysis technique for efficient model compression to arbitrary sizes with minimal quality loss. (3) Inference-time Scaling: A repeated sampling strategy that trades computation for model capacity, enabling smaller models to match larger model quality at inference time. Through these strategies, SANA-1.5 achieves a text-image alignment score of 0.72 on GenEval, which can be further improved to 0.80 through inference scaling, establishing a new SoTA on GenEval benchmark. These innovations enable efficient model scaling across different compute budgets while maintaining high quality, making high-quality image generation more accessible.

This work explores holographic correlators within the frameworks of two-dimensional Boundary Conformal Field Theory (BCFT) and Crosscap Conformal Field Theory (XCFT). Utilizing the AdS/CFT correspondence, we compute stress tensor correlators in BCFT, considering both tensionless and tensionful end-of-the-world (EOW) brane scenarios. We derive recurrence relations for two-point and three-point correlators and examine the impact of non-zero brane tension on correlators. Extending these results, we investigate the holographic duals of XCFTs, presenting explicit scalar and stress tensor correlator computations on projective geometries such as $\mathbb{RP}^2$. Additionally, we analyze stress tensor correlators at a finite cutoff, uncovering deformations to one-point and two-point functions induced by the cutoff. Our findings provide novel insights into the holographic structures of BCFT and XCFT while laying the groundwork for future research into higher-dimensional extensions., Comment: 50 pages, 2 figures