论文检索

Applications · Computer Vision

Duy-Kien Nguyen, Mahmoud Assran, Unnat Jain, Martin R. Oswald, Cees G Snoek, Xinlei Chen

This work does not introduce a new method. Instead, we present an interesting finding that questions the necessity of the inductive bias of locality in modern computer vision architectures. Concretely, we find that vanilla Transformers can operate by directly treating each individual pixel as a token and achieve highly performant results. This is substantially different from the popular design in Vision Transformer, which maintains the inductive bias from ConvNets towards local neighborhoods (e.g., by treating each 16x16 patch as a token). We showcase the effectiveness of pixels-as-tokens across three well-studied computer vision tasks: supervised learning for classification and regression, self-supervised learning via masked autoencoding, and image generation with diffusion models. Although it's computationally less practical to directly operate on individual pixels, we believe the community must be made aware of this surprising piece of knowledge when devising the next generation of neural network architectures for computer vision.

Deep Learning · Self-Supervised Learning

Xinlei Chen, Zhuang Liu, Saining Xie, Kaiming He

In this study, we examine the representation learning abilities of Denoising Diffusion Models (DDM) that were originally purposed for image generation. Our philosophy is to deconstruct a DDM, gradually transforming it into a classical Denoising Autoencoder (DAE). This deconstructive process allows us to explore how various components of modern DDMs influence self-supervised representation learning. We observe that only a very few modern components are critical for learning good representations, while many others are nonessential. Our study ultimately arrives at an approach that is highly simplified and to a large extent resembles a classical DAE. We hope our study will rekindle interest in a family of classical methods within the realm of modern self-supervised learning.

Optimization · Large Scale, Parallel and Distributed

Jialiang Cheng, Ning Gao, Yun Yue, Zhiling Ye, Jiadi Jiang, Jian Sha

Distributed training methods are crucial for large language models (LLMs). However, existing distributed training methods often suffer from communication bottlenecks, stragglers, and limited elasticity, particularly in heterogeneous or large-scale environments. Local SGD methods have been proposed to address these issues, but their effectiveness remains limited to small-scale training due to additional memory overhead and lack of concerns on efficiency and stability. To tackle these issues, we propose EDiT, an innovative Efficient Distributed Training method that combines a tailored Local SGD approach with model sharding techniques to enhance large-scale training efficiency. EDiT performs layer-wise parameter synchronization during forward pass, reducing communication and memory overhead and enabling overlap. Besides, EDiT employs a pseudo gradient penalty strategy to suppress loss spikes, which ensures training stability and improves performance. Additionally, we introduce A-EDiT, a fully asynchronous variant of EDiT that accommodates heterogeneous clusters. Building on EDiT/A-EDiT, we conduct a series of experiments to validate large-scale asynchronous training for LLMs, accompanied by comprehensive analyses. Experimental results demonstrate the superior performance of EDiT/A-EDiT, establishing them as robust solutions for distributed LLM training in diverse computational ecosystems. The code is available at Atorch codebase: https://github.com/intelligent-machine-learning/atorch/tree/main/atorch/local_sgd.

Reinforcement Learning · Multi-agent

Woosung Koh, Wonbeen Oh, Siyeol Kim, Suhin Shin, Hyeongjin Kim, Jaein Jang, Junghyun Lee, Se-Young Yun

Multi-agent reinforcement learning has demonstrated significant potential in addressing complex cooperative tasks across various real-world applications. However, existing MARL approaches often rely on the restrictive assumption that the number of entities (e.g., agents, obstacles) remains constant between training and inference. This overlooks scenarios where entities are dynamically removed or $\textit{added}$ $\textit{during}$ the inference trajectory—a common occurrence in real-world environments like search and rescue missions and dynamic combat situations. In this paper, we tackle the challenge of intra-trajectory dynamic entity composition under zero-shot out-of-domain (OOD) generalization, where such dynamic changes cannot be anticipated beforehand. Our empirical studies reveal that existing MARL methods suffer $\textit{significant}$ performance degradation and increased uncertainty in these scenarios. In response, we propose FlickerFusion, a novel OOD generalization method that acts as a $\textit{universally}$ applicable augmentation technique for MARL backbone methods. FlickerFusion stochastically drops out parts of the observation space, emulating being in-domain when inferenced OOD. The results show that FlickerFusion not only achieves superior inference rewards but also $\textit{uniquely}$ reduces uncertainty vis-à-vis the backbone, compared to existing methods. Benchmarks, implementations, and model weights are organized and open-sourced at $\texttt{\href{flickerfusion305.github.io}{\textbf{flickerfusion305.github.io}}}$, accompanied by ample demo video renderings.

Deep Learning · Attention Mechanisms

Svetha Venkatesh, Kien Do, Hung Le, Truyen Tran, Dung Nguyen

We introduce Pointer-Augmented Neural Memory (PANM), a versatile module designed to enhance neural networks' ability to process symbols and extend their capabilities to longer data sequences. PANM integrates an external neural memory utilizing novel physical addresses and pointer manipulation techniques, emulating human and computer-like symbol processing abilities. PANM facilitates operations like pointer assignment, dereferencing, and arithmetic by explicitly employing physical pointers for memory access. This module can be trained end-to-end on sequence data, empowering various sequential models, from simple recurrent networks to large language models (LLMs). Our experiments showcase PANM's exceptional length extrapolation capabilities and its enhancement of recurrent neural networks in symbol processing tasks, including algorithmic reasoning and Dyck language recognition. PANM enables Transformers to achieve up to 100% generalization accuracy in compositional learning tasks and significantly improves performance in mathematical reasoning, question answering, and machine translation. Notably, the generalization effectiveness scales with stronger backbone models, as evidenced by substantial performance gains when we test LLMs finetuned with PANM for tasks up to 10-100 times longer than the training data.

Kha Pham, Hung Le, Man Ngo, Truyen Tran

While Large Language Models (LLMs) excel at in-context learning (ICL) using just a few demonstrations, their performances are sensitive to demonstration orders. The reasons behind this sensitivity remain poorly understood. In this paper, we investigate the prompt embedding space to bridge the gap between the order sensitivity of ICL with inner workings of decoder-only LLMs, uncovering the clustering property: prompts sharing the first and last demonstrations have closer embeddings, with first-demonstration clustering usually being stronger in practice. We explain this property through extensive theoretical analyses and empirical evidences. Our finding suggests that the positional encoding and the causal attention mask are key contributors to the clustering phenomenon. Leveraging this clustering insight, we introduce Cluster-based Search, a novel method that accelerates the selection and ordering of demonstrations in self-adaptive ICL settings. Our approach substantially decreases the time complexity from factorial to quadratic, saving 92% to nearly 100% execution time while maintaining comparable performance to exhaustive search.

Reinforcement Learning · Function Approximation

Ge Li, Dong Tian, Hongyi Zhou, Xinkai Jiang, Rudolf Lioutikov, Gerhard Neumann

This work introduces Transformer-based Off-Policy Episodic Reinforcement Learning (TOP-ERL), a novel algorithm that enables off-policy updates in the ERL framework. In ERL, policies predict entire action trajectories over multiple time steps instead of single actions at every time step. These trajectories are typically parameterized by trajectory generators such as Movement Primitives (MP), allowing for smooth and efficient exploration over long horizons while capturing high-level temporal correlations. However, ERL methods are often constrained to on-policy frameworks due to the difficulty of evaluating state-action values for entire action sequences, limiting their sample efficiency and preventing the use of more efficient off-policy architectures. TOP-ERL addresses this shortcoming by segmenting long action sequences and estimating the state-action values for each segment using a transformer-based critic architecture alongside an n-step return estimation. These contributions result in efficient and stable training that is reflected in the empirical results conducted on sophisticated robot learning environments. TOP-ERL significantly outperforms state-of-the-art RL methods. Thorough ablation studies additionally show the impact of key design choices on the model performance.

Deep Learning · Generative Models and Autoencoders

Yongjin Yang, Sihyeon Kim, Hojung Jung, Sangmin Bae, SangMook Kim, Se-Young Yun, Kimin Lee

Fine-tuning text-to-image diffusion models with human feedback is an effective method for aligning model behavior with human intentions. However, this alignment process often suffers from slow convergence due to the large size and noise present in human feedback datasets. In this work, we propose FiFA, a novel automated data filtering algorithm designed to enhance the fine-tuning of diffusion models using human feedback datasets with direct preference optimization (DPO). Specifically, our approach selects data by solving an optimization problem to maximize three components: preference margin, text quality, and text diversity. The concept of preference margin is used to identify samples that are highly informative in addressing the noisy nature of feedback dataset, which is calculated using a proxy reward model. Additionally, we incorporate text quality, assessed by large language models to prevent harmful contents, and consider text diversity through a k-nearest neighbor entropy estimator to improve generalization. Finally, we integrate all these components into an optimization process, with approximating the solution by assigning importance score to each data pair and selecting the most important ones. As a result, our method efficiently filters data automatically, without the need for manual intervention, and can be applied to any large-scale dataset. Experimental results show that FiFA significantly enhances training stability and achieves better performance, being preferred by humans 17% more, while using less than 0.5% of the full data and thus 1% of the GPU hours compared to utilizing full human feedback datasets.

Applications · Computer Vision

Anh-Khoa Nguyen Vu, Quoc Truong Truong, Vinh-Tiep Nguyen, Thanh Ngo, Thanh-Toan Do, Tam Nguyen

Recent few-shot object detection (FSOD) methods have focused on augmenting synthetic samples for novel classes, show promising results to the rise of diffusion models. However, the diversity of such datasets is often limited in representativeness because they lack awareness of typical and hard samples, especially in the context of foreground and background relationships. To tackle this issue, we propose a Multi-Perspective Data Augmentation (MPAD) framework. In terms of foreground-foreground relationships, we propose in-context learning for object synthesis (ICOS) with bounding box adjustments to enhance the detail and spatial information of synthetic samples. Inspired by the large margin principle, support samples play a vital role in defining class boundaries. Therefore, we design a Harmonic Prompt Aggregation Scheduler (HPAS) to mix prompt embeddings at each time step of the generation process in diffusion models, producing hard novel samples. For foreground-background relationships, we introduce a Background Proposal method (BAP) to sample typical and hard backgrounds. Extensive experiments on multiple FSOD benchmarks demonstrate the effectiveness of our approach. Our framework significantly outperforms traditional methods, achieving an average increase of $17.5\%$ in nAP50 over the baseline on PASCAL VOC.

Applications · Language, Speech and Dialog

Sungnyun Kim, Sungwoo Cho, Sangmin Bae, Kangwook Jang, Se-Young Yun

Audio-visual speech recognition (AVSR) incorporates auditory and visual modalities to improve recognition accuracy, particularly in noisy environments where audio-only speech systems are insufficient. While previous research has largely addressed audio disruptions, few studies have dealt with visual corruptions, e.g., lip occlusions or blurred videos, which are also detrimental. To address this real-world challenge, we propose CAV2vec, a novel self-supervised speech representation learning framework particularly designed to handle audio-visual joint corruption. CAV2vec employs a self-distillation approach with a corrupted prediction task, where the student model learns to predict clean targets, generated by the teacher model, with corrupted input frames. Specifically, we suggest a unimodal multi-task learning, which distills cross-modal knowledge and aligns the corrupted modalities, by predicting clean audio targets with corrupted videos, and clean video targets with corrupted audios. This strategy mitigates the dispersion in the representation space caused by corrupted modalities, leading to more reliable and robust audio-visual fusion. Our experiments on robust AVSR benchmarks demonstrate that the corrupted representation learning method significantly enhances recognition accuracy across generalized environments involving various types of corruption. Our code is available at https://github.com/sungnyun/cav2vec.

Reinforcement Learning · Multi-agent

Sehyeok Kang, Yongsik Lee, Gahee Kim, Song Chong, Se-Young Yun

Centralized Training and Decentralized Execution (CTDE) is a widely adopted paradigm to solve cooperative multi-agent reinforcement learning (MARL) problems. Despite the successes achieved with CTDE, partial observability still limits cooperation among agents. While previous studies have attempted to overcome this challenge through communication, direct information exchanges could be restricted and introduce additional constraints. Alternatively, if an agent can infer the global information solely from local observations, it can obtain a global view without the need for communication. To this end, we propose the Multi-Agent Masked Auto-Encoder (MA$^2$E), which utilizes the masked auto-encoder architecture to infer the information of other agents from partial observations. By employing masking to learn to reconstruct global information, MA$^2$E serves as an inference module for individual agents within the CTDE framework. MA$^2$E can be easily integrated into existing MARL algorithms and has been experimentally proven to be effective across a wide range of environments and algorithms.

Deep Learning · Large Language Models

Buu Phan, Brandon Amos, Itai Gat, Marton Havasi, Matthew J Muckley, Karen Ullrich

Tokenization is associated with many poorly understood shortcomings in language models (LMs), yet remains an important component for long sequence scaling purposes. This work studies how tokenization impacts model performance by analyzing and comparing the stochastic behavior of tokenized models with their byte-level, or token-free, counterparts. We discover that, even when the two models are statistically equivalent, their predictive distributions over the next byte can be substantially different, a phenomenon we term as ``tokenization bias''. To fully characterize this phenomenon, we introduce the Byte-Token Representation Lemma, a framework that establishes a mapping between the learned token distribution and its equivalent byte-level distribution. From this result, we develop a next-byte sampling algorithm that eliminates tokenization bias without requiring further training or optimization. In other words, this enables zero-shot conversion of tokenized LMs into statistically equivalent token-free ones. We demonstrate its broad applicability with two use cases: fill-in-the-middle (FIM) tasks and model ensembles. In FIM tasks where input prompts may terminate mid-token, leading to out-of-distribution tokenization, our method mitigates performance degradation and achieves 18\% improvement in FIM coding benchmarks, while consistently outperforming the standard token healing fix. For model ensembles where each model employs a distinct vocabulary, our approach enables seamless integration, resulting in improved performance up to 3.7\% over individual models across various standard baselines in reasoning, knowledge, and coding. Code is available at:https: //github.com/facebookresearch/Exact-Byte-Level-Probabilities-from-Tokenized-LMs.

Deep Learning · Generative Models and Autoencoders

Trung X. Pham, Tri Ton, Chang Yoo

We introduce MDSGen, a novel framework for vision-guided open-domain sound generation optimized for model parameter size, memory consumption, and inference speed. This framework incorporates two key innovations: (1) a redundant video feature removal module that filters out unnecessary visual information, and (2) a temporal-aware masking strategy that leverages temporal context for enhanced audio generation accuracy. In contrast to existing resource-heavy Unet-based models, MDSGen employs denoising masked diffusion transformers, facilitating efficient generation without reliance on pre-trained diffusion models. Evaluated on the benchmark VGGSound dataset, our smallest model (5M parameters) achieves 97.9% alignment accuracy, using 172x fewer parameters, 371% less memory, and offering 36x faster inference than the current 860M-parameter state-of-the-art model (93.9% accuracy). The larger model (131M parameters) reaches nearly 99% accuracy while requiring 6.5x fewer parameters. These results highlight the scalability and effectiveness of our approach. The code is available at https://bit.ly/mdsgen.

Deep Learning · Large Language Models

Kaijing Ma, Xeron Du, Yunran Wang, Haoran Zhang, Zhoufutu Wen, Xingwei Qu, Jian Yang, JIAHENG LIU, Minghao Liu, Xiang Yue 等

In this paper, we introduce Knowledge-Orthogonal Reasoning (KOR), a concept aimed at minimizing reliance on domain-specific knowledge, enabling more accurate evaluation of models' reasoning abilities in out-of-distribution settings. Based on this concept, we propose the Knowledge-Orthogonal Reasoning Benchmark (KOR-Bench), encompassing five task categories: Operation, Logic, Cipher, Puzzle, and Counterfactual. KOR-Bench emphasizes models' effectiveness in applying new rule descriptions to solve novel rule-driven questions. O1-Preview and O1-Mini achieve accuracies of 72.88\% and 70.16\%, surpassing Claude-3.5-Sonnet and GPT-4o (58.96\% and 58.00\%), highlighting the effectiveness of KOR-Bench. We perform detailed analyses, identifying bottlenecks in the Cipher task with Stepwise Prompting, where two rounds of Self-Correction yield optimal results. We evaluate performance across three integrated tasks, explore the impact of Tricks on the Puzzle task, and visualize rule-focused attention. Additionally, we conduct an ablation study on dataset size, benchmark correlations, and zero-shot and three-shot "only questions" experiments. KOR-Bench aims to enhance reasoning evaluation and support further research in this area.

Applications · Computer Vision

Muhammad Gohar Javed, chuan guo, Li Cheng, Xingyu Li

Generating realistic 3D human-human interactions from textual descriptions remains a challenging task. Existing approaches, typically based on diffusion models, often produce results lacking realism and fidelity. In this work, we introduce *InterMask*, a novel framework for generating human interactions using collaborative masked modeling in discrete space. InterMask first employs a VQ-VAE to transform each motion sequence into a 2D discrete motion token map. Unlike traditional 1D VQ token maps, it better preserves fine-grained spatio-temporal details and promotes *spatial awareness* within each token. Building on this representation, InterMask utilizes a generative masked modeling framework to collaboratively model the tokens of two interacting individuals. This is achieved by employing a transformer architecture specifically designed to capture complex spatio-temporal inter-dependencies. During training, it randomly masks the motion tokens of both individuals and learns to predict them. For inference, starting from fully masked sequences, it progressively fills in the tokens for both individuals. With its enhanced motion representation, dedicated architecture, and effective learning strategy, InterMask achieves state-of-the-art results, producing high-fidelity and diverse human interactions. It outperforms previous methods, achieving an FID of $5.154$ (vs $5.535$ of in2IN) on the InterHuman dataset and $0.399$ (vs $5.207$ of InterGen) on the InterX dataset. Additionally, InterMask seamlessly supports reaction generation without the need for model redesign or fine-tuning.

Social Aspects · Privacy-preserving Statistics and Machine Learning

Yunpeng Zhao, Jie Zhang

As synthetic data becomes increasingly popular in machine learning tasks, numerous methods---without formal differential privacy guarantees---use synthetic data for training. These methods often claim, either explicitly or implicitly, to protect the privacy of the original training data.In this work, we explore four different training paradigms: coreset selection, dataset distillation, data-free knowledge distillation, and synthetic data generated from diffusion models. While all these methods utilize synthetic data for training, they lead to vastly different conclusions regarding privacy preservation. We caution that empirical approaches to preserving data privacy require careful and rigorous evaluation; otherwise, they risk providing a false sense of privacy.

Applications · Computer Vision

Mohamed el amine Boudjoghra, Angela Dai, Jean Lahoud, Hisham Cholakkal, Rao Anwer, Salman Khan, Fahad Khan

Recent works on open-vocabulary 3D instance segmentation show strong promise but at the cost of slow inference speed and high computation requirements. This high computation cost is typically due to their heavy reliance on aggregated clip features from multi-view, which require computationally expensive 2D foundation models like Segment Anything (SAM) and CLIP. Consequently, this hampers their applicability in many real-world applications that require both fast and accurate predictions. To this end, we propose a novel open-vocabulary 3D instance segmentation approach, named Open-YOLO 3D, that efficiently leverages only 2D object detection from multi-view RGB images for open-vocabulary 3D instance segmentation. We demonstrate that our proposed Multi-View Prompt Distribution (MVPDist) method makes use of multi-view information to account for misclassification from the object detector to predict a reliable label for 3D instance masks. Furthermore, since projections of 3D object instances are already contained within the 2D bounding boxes, we show that our proposed low granularity label maps, which require only a 2D object detector to construct, are sufficient and very fast to predict prompt IDs for 3D instance masks when used with our proposed MVPDist. We validate our Open-YOLO 3D on two benchmarks, ScanNet200 and Replica, under two scenarios: (i) with ground truth masks, where labels are required for given object proposals, and (ii) with class-agnostic 3D proposals generated from a 3D proposal network. Our Open-YOLO 3D achieves state-of-the-art performance on both datasets while obtaining up to $\sim$16$\times$ speedup compared to the best existing method in literature. On ScanNet200 val. set, our Open-YOLO 3D achieves mean average precision (mAP) of 24.7% while operating at 22 seconds per scene. github.com/aminebdj/OpenYOLO3D

Deep Learning · Generative Models and Autoencoders

Minh Quan Dao, Khanh Doan, Di Liu, Trung Le, Dimitris Metaxas

Consistency models are a new family of generative models capable of producing high-quality samples in either a single step or multiple steps. Recently, consistency models have demonstrated impressive performance, achieving results on par with diffusion models in the pixel space. However, the success of scaling consistency training to large-scale datasets, particularly for text-to-image and video generation tasks, is determined by performance in the latent space. In this work, we analyze the statistical differences between pixel and latent spaces, discovering that latent data often contains highly impulsive outliers, which significantly degrade the performance of iCT in the latent space. To address this, we replace Pseudo-Huber losses with Cauchy losses, effectively mitigating the impact of outliers. Additionally, we introduce a diffusion loss at early timesteps and employ optimal transport (OT) coupling to further enhance performance. Lastly, we introduce the adaptive scaling-$c$ scheduler to manage the robust training process and adopt Non-scaling LayerNorm in the architecture to better capture the statistics of the features and reduce outlier impact. With these strategies, we successfully train latent consistency models capable of high-quality sampling with one or two steps, significantly narrowing the performance gap between latent consistency and diffusion models. The implementation is released here: \url{https://github.com/quandao10/sLCT/}

Deep Learning · Other Representation Learning

Yuning Cui, Syed Waqas Zamir, Salman Khan, Alois Knoll, Mubarak Shah, Fahad Khan

In the image acquisition process, various forms of degradation, including noise, blur, haze, and rain, are frequently introduced. These degradations typically arise from the inherent limitations of cameras or unfavorable ambient conditions. To recover clean images from their degraded versions, numerous specialized restoration methods have been developed, each targeting a specific type of degradation. Recently, all-in-one algorithms have garnered significant attention by addressing different types of degradations within a single model without requiring the prior information of the input degradation type. However, most methods purely operate in the spatial domain and do not delve into the distinct frequency variations inherent to different degradation types. To address this gap, we propose an adaptive all-in-one image restoration network based on frequency mining and modulation. Our approach is motivated by the observation that different degradation types impact the image content on different frequency subbands, thereby requiring different treatments for each restoration task. Specifically, we first mine low- and high-frequency information from the input features, guided by the adaptively decoupled spectra of the degraded image. The extracted features are then modulated by a bidirectional operator to facilitate interactions between different frequency components. Finally, the modulated features are merged into the original input for a progressively guided restoration. With this approach, the model achieves adaptive reconstruction by accentuating the informative frequency subbands according to different input degradations. Extensive experiments demonstrate that the proposed method, AdaIR, achieves state-of-the-art performance on different image restoration tasks, including image denoising, dehazing, deraining, motion deblurring, and low-light image enhancement. The code is available at https://github.com/c-yn/AdaIR.

Applications · Computer Vision

Mohamed el amine Boudjoghra, Angela Dai, Jean Lahoud, Hisham Cholakkal, Rao Anwer, Salman Khan, Fahad Khan

Recent works on open-vocabulary 3D instance segmentation show strong promise but at the cost of slow inference speed and high computation requirements. This high computation cost is typically due to their heavy reliance on aggregated clip features from multi-view, which require computationally expensive 2D foundation models like Segment Anything (SAM) and CLIP. Consequently, this hampers their applicability in many real-world applications that require both fast and accurate predictions. To this end, we propose a novel open-vocabulary 3D instance segmentation approach, named Open-YOLO 3D, that efficiently leverages only 2D object detection from multi-view RGB images for open-vocabulary 3D instance segmentation. We demonstrate that our proposed Multi-View Prompt Distribution (MVPDist) method makes use of multi-view information to account for misclassification from the object detector to predict a reliable label for 3D instance masks. Furthermore, since projections of 3D object instances are already contained within the 2D bounding boxes, we show that our proposed low granularity label maps, which require only a 2D object detector to construct, are sufficient and very fast to predict prompt IDs for 3D instance masks when used with our proposed MVPDist. We validate our Open-YOLO 3D on two benchmarks, ScanNet200 and Replica, under two scenarios: (i) with ground truth masks, where labels are required for given object proposals, and (ii) with class-agnostic 3D proposals generated from a 3D proposal network. Our Open-YOLO 3D achieves state-of-the-art performance on both datasets while obtaining up to $\sim$16$\times$ speedup compared to the best existing method in literature. On ScanNet200 val. set, our Open-YOLO 3D achieves mean average precision (mAP) of 24.7% while operating at 22 seconds per scene. github.com/aminebdj/OpenYOLO3D