통합 멀티모달 모델에서 문맥 내 학습(ICL)의 비단조적 스케일링 문제를 진단하기 위해, 지각부터 고차 식별까지 시연 예제의 기능적 역할을 6단계로 분류하는 능력 지향 분류체계(Capability-Oriented Taxonomy)를 제안하고 이를 기반으로 15개 서브태스크에 걸친 대규모 코퍼스 UniICL-760K와 벤치마크 UniICL-Bench를 구축하였다. 경량 플러그앤플레이 모듈인 CAPM(Context-Adaptive Prototype Modulator)을 추가 안정화 장치로 결합한 결과, 더 많은 파라미터를 가진 멀티모달 대규모 언어 모델(MLLM) 베이스라인을 대부분의 이해 태스크에서 능가하였다.
특화 도메인에 맞춰 파인튜닝(fine-tuning)된 대규모 언어 모델(LLM)은 매우 중요한 고영향(high-impact) 응용 분야를 나타냅니다. 추론(inference) 시점 정렬(alignment)은 상당한 연산 자원을 요구하지 않고도 특화 파인튜닝으로 인해 저하된 안전성(safety)을 향상시키며, 사용하기 쉬운 플러그 앤 플레이(plug-and-play) 솔루션으로서 파인튜닝 기반 방식을 보완합니다. 그러나 기존 추론 시점 방식들은 도메인 역량을 훼손하지 않으면서 안전성을 안정적으로 개선하는 데 실패합니다. 우리는 근본 원인을 상호보완적 전문성의 직교성(complementary expertise orthogonality)으로 규명합니다. 특화된 베이스 모델과 일반 도메인 가이던스 모델은 서로 직교하는 역량을 가지고 있어, 가이던스 신호가 특화 생성에 비신뢰적이게 됩니다. 이는 주로 정지 토큰(stop token) 간섭으로 나타나는데, 가이던스 모델의 생성 지속 경향이 베이스 모델의 중단 결정을 무시하여 정답이 가이던스로 유도된 생성 지속 아래 묻히게 만듭니다. 이 문제를 해결하기 위해, 우리는 모든 패밀리의 가이던스 모델에서 추출한 안전성 방향을 활용하여 베이스 모델의 은닉 표상(hidden representation)을 스티어링함으로써 토큰 수준의 구조적 한계를 완전히 회피하는 추론 시점 정렬 기법인 CREST를 제안합니다. CREST는 특화로 인해 약화된 안전성을 개선하는 동시에 도메인 특화 역량과 이미 잘 정렬된 모델의 안전성을 모두 보존하며, 안전성 벤치마크(benchmark)에서 베이스라인 대비 최대 22.2\% 뛰어난 성능을 보입니다. 코드 사용 링크: https://github.com/DecayingSeart/CREST.
Interactive visualizations support conceptual understanding in undergraduate mathematics, but building them has required programming expertise most instructors lack. Using a design-based research approach, we develop, deploy, and evaluate a six-phase workflow (Foundation, Customization, Mathematical Depth, Application, Accessibility, Pedagogical Control) that uses generative AI to build WCAG~2.2 Level~AA compliant visualizations without programming. The six phases structure every prompt, scaffold the AI's code generation, and define where human verification is applied. We ask whether the structure reliably yields correct and accessible tools, whether it runs both backward (reverse-engineering prompts from a finished tool) and forward (generating a tool from a plain-language idea), and what verification each phase requires. Across four deployed tools spanning calculus, multivariable calculus, and differential equations, we evaluate mathematical correctness against closed forms, accessibility through automated and manual screen-reader testing, and the errors that recurred. The structure produces structurally complete first-pass tools, but human verification remains mandatory at every phase: each output must be checked for mathematical correctness, accessibility, and pedagogical fit before the next phase begins. The workflow is platform-independent and serves both instructors and students.
시각 모달리티, 즉 이미지는 다중모달 엔티티 정렬(MMEA)에서 핵심적인 역할을 한다. 기존 접근법은 종종 이미지를 다른 모달리티와 직접 융합하여 서로 다른 엔티티를 정렬한다. 단순하지만, 이러한 전략은 이미지에 존재할 수 있는 노이즈와 해당 엔티티와의 의미적 불일치를 간과하여, 차선의 융합과 성능 저하를 초래한다. 이에 대응하여, 우리는 시각 신뢰도를 평가하고 견고한 엔티티 정렬을 위해 신뢰할 수 없는 시각 표현을 적응적으로 개선하는 MMEA용 신규 신뢰도 인지 프레임워크(RA-MMEA)를 제안한다. 핵심은 의존성 인지 시각 신뢰도 예측(DA-VRP)과 안정성 정규화 시각 임베딩 생성(SR-VEG)을 포함한 두 모듈에 있다. 전자는 엔티티 내 다중모달 의존성을 활용하여 이미지의 신뢰도를 추정하는 것을 목표로 하고, 후자는 다중모달 융합을 위해 텍스트 모달리티에 인코딩된 의미에 조건화된 대안적 시각 표현을 생성하는 데 초점을 둔다. 현재 방법들과 비교할 때, RA-MMEA는 모달리티 융합을 위한 더 신뢰할 수 있는 시각 표현을 가능하게 하여 성능을 향상시킨다. 광범위한 실험에서 RA-MMEA는 최신 최고 성능(state-of-the-art) 결과를 달성하여, 엔티티 정렬을 위한 신뢰할 수 있는 시각 모달리티의 중요성과 RA-MMEA의 효과를 검증한다. 코드와 결과는 공개될 예정이다.
This paper presents, evaluates, and discusses a new software tool to automatically build Dynamic Bayesian Networks (DBNs) from ordinary differential equations (ODEs) entered by the user. The DBNs generated from ODE models can handle both data uncertainty and model uncertainty in a principled manner. The application, named PROFET, can be used for temporal data mining with noisy or missing variables. It enables automatic re-estimation of model parameters using temporal evidence in the form of data streams. For temporal inference, PROFET includes both standard fixed time step particle filtering and its extension, adaptive-time particle filtering algorithms. Adaptive-time particle filtering enables the DBN to automatically adapt its time step length to match the dynamics of the model. We demonstrate PROFET's functionality by using it to infer the model variables by estimating the model parameters of four benchmark ODE systems. From the generation of the DBN model to temporal inference, the entire process is automated and is delivered as an open-source platform-independent software application with a comprehensive user interface. PROFET is released under the Apache License 2.0. Its source code, executable and documentation are available at http:://profet.it.nuigalway.ie.
Attention mechanism has been used as an important component across Vision-and-Language(VL) tasks in order to bridge the semantic gap between visual and textual features. While attention has been widely used in VL tasks, it has not been examined the capability of different attention alignment calculation in bridging the semantic gap between visual and textual clues. In this research, we conduct a comprehensive analysis on understanding the role of attention alignment by looking into the attention score calculation methods and check how it actually represents the visual region's and textual token's significance for the global assessment. We also analyse the conditions which attention score calculation mechanism would be more (or less) interpretable, and which may impact the model performance on three different VL tasks, including visual question answering, text-to-image generation, text-and-image matching (both sentence and image retrieval). Our analysis is the first of its kind and provides useful insights of the importance of each attention alignment score calculation when applied at the training phase of VL tasks, commonly ignored in attention-based cross modal models, and/or pretrained models. Our code is available at: https://github.com/adlnlp/Attention_VL
With the continuous development of deep learning in the field of image generation models, a large number of vivid forged faces have been generated and spread on the Internet. These high-authenticity artifacts could grow into a threat to society security. Existing face forgery detection methods directly utilize the obtained public shared or centralized data for training but ignore the personal privacy and security issues when personal data couldn't be centralizedly shared in real-world scenarios. Additionally, different distributions caused by diverse artifact types would further bring adverse influences on the forgery detection task. To solve the mentioned problems, the paper proposes a novel generalized residual Federated learning for face Forgery detection (FedForgery). The designed variational autoencoder aims to learn robust discriminative residual feature maps to detect forgery faces (with diverse or even unknown artifact types). Furthermore, the general federated learning strategy is introduced to construct distributed detection model trained collaboratively with multiple local decentralized devices, which could further boost the representation generalization. Experiments conducted on publicly available face forgery detection datasets prove the superior performance of the proposed FedForgery. The designed novel generalized face forgery detection protocols and source code would be publicly available.
We propose an end-to-end one-step person search approach with learnable proposals, named LEAPS. Given a set of sparse and learnable proposals, LEAPS employs a dynamic person search head to directly perform person detection and corresponding re-id feature generation without non-maximum suppression post-processing. The dynamic person search head comprises a detection head and a novel flexible re-id head. Our flexible re-id head first employs a dynamic region-of-interest (RoI) operation to extract discriminative RoI features of the proposals. Then, it generates re-id features using a plain and a hierarchical interaction re-id module. To better guide discriminative re-id feature learning, we introduce a diverse re-id sample matching strategy, instead of bipartite matching in detection head. Comprehensive experiments reveal the benefit of the proposed LEAPS, achieving a favorable performance on two public person search benchmarks: CUHK-SYSU and PRW. When using the same ResNet50 backbone, our LEAPS obtains a mAP score of 55.0%, outperforming the best reported results in literature by 1.7%, while achieving around a two-fold speedup on the challenging PRW dataset. Our source code and models will be released.
With the emergence of image super-resolution (SR) algorithm, how to blindly evaluate the quality of super-resolution images has become an urgent task. However, existing blind SR image quality assessment (IQA) metrics merely focus on visual characteristics of super-resolution images, ignoring the available scale information. In this paper, we reveal that the scale factor has a statistically significant impact on subjective quality scores of SR images, indicating that the scale information can be used to guide the task of blind SR IQA. Motivated by this, we propose a scale guided hypernetwork framework that evaluates SR image quality in a scale-adaptive manner. Specifically, the blind SR IQA procedure is divided into three stages, i.e., content perception, evaluation rule generation, and quality prediction. After content perception, a hypernetwork generates the evaluation rule used in quality prediction based on the scale factor of the SR image. We apply the proposed scale guided hypernetwork framework to existing representative blind IQA metrics, and experimental results show that the proposed framework not only boosts the performance of these IQA metrics but also enhances their generalization abilities. Source code will be available at https://github.com/JunFu1995/SGH.
Recent development of large language models (LLMs), such as ChatGPT has been widely applied to a wide range of software engineering tasks. Many papers have reported their analysis on the potential advantages and limitations of ChatGPT for writing code, summarization, text generation, etc. However, the analysis of the current state of ChatGPT for log processing has received little attention. Logs generated by large-scale software systems are complex and hard to understand. Despite their complexity, they provide crucial information for subject matter experts to understand the system status and diagnose problems of the systems. In this paper, we investigate the current capabilities of ChatGPT to perform several interesting tasks on log data, while also trying to identify its main shortcomings. Our findings show that the performance of the current version of ChatGPT for log processing is limited, with a lack of consistency in responses and scalability issues. We also outline our views on how we perceive the role of LLMs in the log processing discipline and possible next steps to improve the current capabilities of ChatGPT and the future LLMs in this area. We believe our work can contribute to future academic research to address the identified issues.
Speech-driven 3D facial animation aims at generating facial movements that are synchronized with the driving speech, which has been widely explored recently. Existing works mostly neglect the person-specific talking style in generation, including facial expression and head pose styles. Several works intend to capture the personalities by fine-tuning modules. However, limited training data leads to the lack of vividness. In this work, we propose AdaMesh, a novel adaptive speech-driven facial animation approach, which learns the personalized talking style from a reference video of about 10 seconds and generates vivid facial expressions and head poses. Specifically, we propose mixture-of-low-rank adaptation (MoLoRA) to fine-tune the expression adapter, which efficiently captures the facial expression style. For the personalized pose style, we propose a pose adapter by building a discrete pose prior and retrieving the appropriate style embedding with a semantic-aware pose style matrix without fine-tuning. Extensive experimental results show that our approach outperforms state-of-the-art methods, preserves the talking style in the reference video, and generates vivid facial animation. The supplementary video and code will be available at https://adamesh.github.io.
Structured dropout approaches, such as attention dropout and DropHead, have been investigated to regularize the multi-head attention mechanism in Transformers. In this paper, we propose a new regularization scheme based on token-level rather than structure-level to reduce overfitting. Specifically, we devise a novel Token-Level Masking (TLM) training strategy for Transformers to regularize the connections of self-attention, which consists of two masking techniques that are effective and easy to implement. The underlying idea is to manipulate the connections between tokens in the multi-head attention via masking, where the networks are forced to exploit partial neighbors' information to produce a meaningful representation. The generality and effectiveness of TLM are thoroughly evaluated via extensive experiments on 4 diversified NLP tasks across 18 datasets, including natural language understanding benchmark GLUE, ChineseGLUE, Chinese Grammatical Error Correction, and data-to-text generation. The results indicate that TLM can consistently outperform attention dropout and DropHead, e.g., it increases by 0.5 points relative to DropHead with BERT-large on GLUE. Moreover, TLM can establish a new record on the data-to-text benchmark Rotowire (18.93 BLEU). Our code will be publicly available at https://github.com/Young1993/tlm.
Deep learning is widely applied in computer-aided pathological diagnosis, which alleviates the pathologist workload and provide timely clinical analysis. However, most models generally require large-scale annotated data for training, which faces challenges due to the sampling and annotation scarcity in pathological images. The rapid developing generative models shows potential to generate more training samples from recent studies. However, they also struggle in generalization diversity with limited training data, incapable of generating effective samples. Inspired by the pathological transitions between different stages, we propose an adaptive depth-controlled diffusion (ADD) network to generate pathological progressive images for effective data augmentation. This novel approach roots in domain migration, where a hybrid attention strategy guides the bidirectional diffusion, blending local and global attention priorities. With feature measuring, the adaptive depth-controlled strategy ensures the migration and maintains locational similarity in simulating the pathological feature transition. Based on tiny training set (samples less than 500), the ADD yields cross-domain progressive images with corresponding soft-labels. Experiments on two datasets suggest significant improvements in generation diversity, and the effectiveness with generated progressive samples are highlighted in downstream classifications. The code is available at https://github.com/Rowerliu/ADD.
The demand for efficient machine learning (ML) accelerators is growing rapidly, driving the development of novel computing concepts such as resistive random access memory (RRAM)-based tiled computing-in-memory (CIM) architectures. CIM allows to compute within the memory unit, resulting in faster data processing and reduced power consumption. Efficient compiler algorithms are essential to exploit the potential of tiled CIM architectures. While conventional ML compilers focus on code generation for CPUs, GPUs, and other von Neumann architectures, adaptations are needed to cover CIM architectures. Cross-layer scheduling is a promising approach, as it enhances the utilization of CIM cores, thereby accelerating computations. Although similar concepts are implicitly used in previous work, there is a lack of clear and quantifiable algorithmic definitions for cross-layer scheduling for tiled CIM architectures. To close this gap, we present CLSA-CIM, a cross-layer scheduling algorithm for tiled CIM architectures. We integrate CLSA-CIM with existing weight-mapping strategies and compare performance against state-of-the-art (SOTA) scheduling algorithms. CLSA-CIM improves the utilization by up to 17.9 x , resulting in an overall speedup increase of up to 29.2 x compared to SOTA.
Multi-modal 3D object detection models for automated driving have demonstrated exceptional performance on computer vision benchmarks like nuScenes. However, their reliance on densely sampled LiDAR point clouds and meticulously calibrated sensor arrays poses challenges for real-world applications. Issues such as sensor misalignment, miscalibration, and disparate sampling frequencies lead to spatial and temporal misalignment in data from LiDAR and cameras. Additionally, the integrity of LiDAR and camera data is often compromised by adverse environmental conditions such as inclement weather, leading to occlusions and noise interference. To address this challenge, we introduce MultiCorrupt, a comprehensive benchmark designed to evaluate the robustness of multi-modal 3D object detectors against ten distinct types of corruptions. We evaluate five state-of-the-art multi-modal detectors on MultiCorrupt and analyze their performance in terms of their resistance ability. Our results show that existing methods exhibit varying degrees of robustness depending on the type of corruption and their fusion strategy. We provide insights into which multi-modal design choices make such models robust against certain perturbations. The dataset generation code and benchmark are open-sourced at https://github.com/ika-rwth-aachen/MultiCorrupt.
Spiking Neural Networks (SNNs) have emerged as a promising third generation of neural networks, offering unique characteristics such as binary outputs, high sparsity, and biological plausibility. However, the lack of effective learning algorithms remains a challenge for SNNs. For instance, while converting artificial neural networks (ANNs) to SNNs circumvents the need for direct training of SNNs, it encounters issues related to conversion errors and high inference time delays. In order to reduce or even eliminate conversion errors while decreasing inference time-steps, we have introduced a novel type of neuron called Group Neurons (GNs). One GN is composed of multiple Integrate-and-Fire (IF) neurons as members, and its neural dynamics are meticulously designed. Based on GNs, we have optimized the traditional ANN-SNN conversion framework. Specifically, we replace the IF neurons in the SNNs obtained by the traditional conversion framework with GNs. The resulting SNNs, which utilize GNs, are capable of achieving accuracy levels comparable to ANNs even within extremely short inference time-steps. The experiments on CIFAR10, CIFAR100, and ImageNet datasets demonstrate the superiority of the proposed methods in terms of both inference accuracy and latency. Code is available at https://github.com/Lyu6PosHao/ANN2SNN_GN.
As the adoption of large language models increases and the need for per-user or per-task model customization grows, the parameter-efficient fine-tuning (PEFT) methods, such as low-rank adaptation (LoRA) and its variants, incur substantial storage and transmission costs. To further reduce stored parameters, we introduce a "divide-and-share" paradigm that breaks the barriers of low-rank decomposition across matrix dimensions, modules, and layers by sharing parameters globally via a vector bank. As an instantiation of the paradigm to LoRA, our proposed VB-LoRA composites all the low-rank matrices of LoRA from a shared vector bank with a differentiable top-k admixture module. VB-LoRA achieves extreme parameter efficiency while maintaining comparable or better performance compared to state-of-the-art PEFT methods. Extensive experiments demonstrate the effectiveness of VB-LoRA on natural language understanding, natural language generation, instruction tuning, and mathematical reasoning tasks. When fine-tuning the Llama2-13B model, VB-LoRA only uses 0.4% of LoRA's stored parameters, yet achieves superior results. Our source code is available at https://github.com/leo-yangli/VB-LoRA. This method has been merged into the Hugging Face PEFT package.
Positron emission tomography (PET) is a well-established functional imaging technique for diagnosing brain disorders. However, PET's high costs and radiation exposure limit its widespread use. In contrast, magnetic resonance imaging (MRI) does not have these limitations. Although it also captures neurodegenerative changes, MRI is a less sensitive diagnostic tool than PET. To close this gap, we aim to generate synthetic PET from MRI. Herewith, we introduce PASTA, a novel pathology-aware image translation framework based on conditional diffusion models. Compared to the state-of-the-art methods, PASTA excels in preserving both structural and pathological details in the target modality, which is achieved through its highly interactive dual-arm architecture and multi-modal condition integration. A cycle exchange consistency and volumetric generation strategy elevate PASTA's capability to produce high-quality 3D PET scans. Our qualitative and quantitative results confirm that the synthesized PET scans from PASTA not only reach the best quantitative scores but also preserve the pathology correctly. For Alzheimer's classification, the performance of synthesized scans improves over MRI by 4%, almost reaching the performance of actual PET. Code is available at https://github.com/ai-med/PASTA.
Automated Machine Learning (AutoML) offers a promising approach to streamline the training of machine learning models. However, existing AutoML frameworks are often limited to unimodal scenarios and require extensive manual configuration. Recent advancements in Large Language Models (LLMs) have showcased their exceptional abilities in reasoning, interaction, and code generation, presenting an opportunity to develop a more automated and user-friendly framework. To this end, we introduce AutoM3L, an innovative Automated Multimodal Machine Learning framework that leverages LLMs as controllers to automatically construct multimodal training pipelines. AutoM3L comprehends data modalities and selects appropriate models based on user requirements, providing automation and interactivity. By eliminating the need for manual feature engineering and hyperparameter optimization, our framework simplifies user engagement and enables customization through directives, addressing the limitations of previous rule-based AutoML approaches. We evaluate the performance of AutoM3L on six diverse multimodal datasets spanning classification, regression, and retrieval tasks, as well as a comprehensive set of unimodal datasets. The results demonstrate that AutoM3L achieves competitive or superior performance compared to traditional rule-based AutoML methods. Furthermore, a user study highlights the user-friendliness and usability of our framework, compared to the rule-based AutoML methods.
The focus of this paper is on 3D motion editing. Given a 3D human motion and a textual description of the desired modification, our goal is to generate an edited motion as described by the text. The key challenges include the scarcity of training data and the need to design a model that accurately edits the source motion. In this paper, we address both challenges. We propose a methodology to semi-automatically collect a dataset of triplets comprising (i) a source motion, (ii) a target motion, and (iii) an edit text, introducing the new MotionFix dataset. Access to this data allows us to train a conditional diffusion model, TMED, that takes both the source motion and the edit text as input. We develop several baselines to evaluate our model, comparing it against models trained solely on text-motion pair datasets, and demonstrate the superior performance of our model trained on triplets. We also introduce new retrieval-based metrics for motion editing, establishing a benchmark on the evaluation set of MotionFix. Our results are promising, paving the way for further research in fine-grained motion generation. Code, models, and data are available at https://motionfix.is.tue.mpg.de/ .