Diffusion models have achieved great success in image generation, with the backbone evolving from U-Net to Vision Transformers. However, the computational cost of Transformers is quadratic to the number of tokens, leading to significant challenges when dealing with high-resolution images. In this work, we propose Diffusion Mamba (DiM), which combines the efficiency of Mamba, a sequence model based on State Space Models (SSM), with the expressive power of diffusion models for efficient high-resolution image synthesis. To address the challenge that Mamba cannot generalize to 2D signals, we make several architecture designs including multi-directional scans, learnable padding tokens at the end of each row and column, and lightweight local feature enhancement. Our DiM architecture achieves inference-time efficiency for high-resolution images. In addition, to further improve training efficiency for high-resolution image generation with DiM, we investigate "weak-to-strong" training strategy that pretrains DiM on low-resolution images ($256\times 256$) and then finetune it on high-resolution images ($512 \times 512$). We further explore training-free upsampling strategies to enable the model to generate higher-resolution images (e.g., $1024\times 1024$ and $1536\times 1536$) without further fine-tuning. Experiments demonstrate the effectiveness and efficiency of our DiM. The code of our work is available here: {\url{https://github.com/tyshiwo1/DiM-DiffusionMamba/}}.
Large language models applied to vast biological datasets have the potential to transform biology by uncovering disease mechanisms and accelerating drug development. However, current models are often siloed, trained separately on small-molecules, proteins, or transcriptomic data, limiting their ability to capture complex, multi-modal interactions. Effective drug discovery requires computational tools that integrate multiple biological entities while supporting prediction and generation, a challenge existing models struggle to address. For this purpose, we present MAMMAL - Molecular Aligned Multi-Modal Architecture and Language - a versatile method applied to create a multi-task foundation model that learns from large-scale biological datasets across diverse modalities, including proteins, small-molecules, and omics. MAMMAL's structured prompt syntax supports classification, regression, and generation tasks while handling token and scalar inputs and outputs. Evaluated on eleven diverse downstream tasks, it reaches a new state of the art (SOTA) in nine tasks and is comparable to SOTA in two tasks, all within a unified architecture, unlike prior task-specific models. Additionally, we explored Alphafold 3 binding prediction capabilities on antibody-antigen and nanobody-antigen complexes showing significantly better classification performance of MAMMAL in 3 out of 4 targets. The model code and pretrained weights are publicly available at https://github.com/BiomedSciAI/biomed-multi-alignment and https://huggingface.co/ibm/biomed.omics.bl.sm.ma-ted-458m
Multimodal Retrieval Augmented Generation (mRAG) plays an important role in mitigating the "hallucination" issue inherent in multimodal large language models (MLLMs). Although promising, existing heuristic mRAGs typically predefined fixed retrieval processes, which causes two issues: (1) Non-adaptive Retrieval Queries. (2) Overloaded Retrieval Queries. However, these flaws cannot be adequately reflected by current knowledge-seeking visual question answering (VQA) datasets, since the most required knowledge can be readily obtained with a standard two-step retrieval. To bridge the dataset gap, we first construct Dyn-VQA dataset, consisting of three types of "dynamic" questions, which require complex knowledge retrieval strategies variable in query, tool, and time: (1) Questions with rapidly changing answers. (2) Questions requiring multi-modal knowledge. (3) Multi-hop questions. Experiments on Dyn-VQA reveal that existing heuristic mRAGs struggle to provide sufficient and precisely relevant knowledge for dynamic questions due to their rigid retrieval processes. Hence, we further propose the first self-adaptive planning agent for multimodal retrieval, OmniSearch. The underlying idea is to emulate the human behavior in question solution which dynamically decomposes complex multimodal questions into sub-question chains with retrieval action. Extensive experiments prove the effectiveness of our OmniSearch, also provide direction for advancing mRAG. The code and dataset will be open-sourced at https://github.com/Alibaba-NLP/OmniSearch.
We address the task of multi-view image-to-image translation for person image generation. The goal is to synthesize photo-realistic multi-view images with pose-consistency across all views. Our proposed end-to-end framework is based on a joint learning of multiple unpaired image-to-image translation models, one per camera viewpoint. The joint learning is imposed by constraints on the shared 3D human pose in order to encourage the 2D pose projections in all views to be consistent. Experimental results on the CMU-Panoptic dataset demonstrate the effectiveness of the suggested framework in generating photo-realistic images of persons with new poses that are more consistent across all views in comparison to a standard Image-to-Image baseline. The code is available at: https://github.com/sony-si/MultiView-Img2Img
We present in this paper an informed single-channel dereverberation method based on conditional generation with diffusion models. With knowledge of the room impulse response, the anechoic utterance is generated via reverse diffusion using a measurement consistency criterion coupled with a neural network that represents the clean speech prior. The proposed approach is largely more robust to measurement noise compared to a state-of-the-art informed single-channel dereverberation method, especially for non-stationary noise. Furthermore, we compare to other blind dereverberation methods using diffusion models and show superiority of the proposed approach for large reverberation times. We motivate the presented algorithm by introducing an extension for blind dereverberation allowing joint estimation of the room impulse response and anechoic speech. Audio samples and code can be found online (https://uhh.de/inf-sp-derev-dps).
Generative Flow Networks (GFlowNets) are amortized sampling methods that learn a distribution over discrete objects proportional to their rewards. GFlowNets exhibit a remarkable ability to generate diverse samples, yet occasionally struggle to consistently produce samples with high rewards due to over-exploration on wide sample space. This paper proposes to train GFlowNets with local search, which focuses on exploiting high-rewarded sample space to resolve this issue. Our main idea is to explore the local neighborhood via backtracking and reconstruction guided by backward and forward policies, respectively. This allows biasing the samples toward high-reward solutions, which is not possible for a typical GFlowNet solution generation scheme, which uses the forward policy to generate the solution from scratch. Extensive experiments demonstrate a remarkable performance improvement in several biochemical tasks. Source code is available: \url{https://github.com/dbsxodud-11/ls_gfn}.
Counterfactual reasoning is pivotal in human cognition and especially important for providing explanations and making decisions. While Judea Pearl's influential approach is theoretically elegant, its generation of a counterfactual scenario often requires too much deviation from the observed scenarios to be feasible, as we show using simple examples. To mitigate this difficulty, we propose a framework of \emph{natural counterfactuals} and a method for generating counterfactuals that are more feasible with respect to the actual data distribution. Our methodology incorporates a certain amount of backtracking when needed, allowing changes in causally preceding variables to minimize deviations from realistic scenarios. Specifically, we introduce a novel optimization framework that permits but also controls the extent of backtracking with a naturalness criterion. Empirical experiments demonstrate the effectiveness of our method. The code is available at https://github.com/GuangyuanHao/natural_counterfactuals.
In recent years, advancements in representation learning and language models have propelled Automated Captioning (AC) to new heights, enabling the generation of human-level descriptions. Leveraging these advancements, we propose AVCap, an Audio-Visual Captioning framework, a simple yet powerful baseline approach applicable to audio-visual captioning. AVCap utilizes audio-visual features as text tokens, which has many advantages not only in performance but also in the extensibility and scalability of the model. AVCap is designed around three pivotal dimensions: the exploration of optimal audio-visual encoder architectures, the adaptation of pre-trained models according to the characteristics of generated text, and the investigation into the efficacy of modality fusion in captioning. Our method outperforms existing audio-visual captioning methods across all metrics and the code is available on https://github.com/JongSuk1/AVCap
Nonlinear activation functions are pivotal to the success of deep neural nets, and choosing the appropriate activation function can significantly affect their performance. Most networks use fixed activation functions (e.g., ReLU, GELU, etc.), and this choice might limit their expressiveness. Furthermore, different layers may benefit from diverse activation functions. Consequently, there has been a growing interest in trainable activation functions. In this paper, we introduce DiTAC, a trainable highly-expressive activation function based on an efficient diffeomorphic transformation (called CPAB). Despite introducing only a negligible number of trainable parameters, DiTAC enhances model expressiveness and performance, often yielding substantial improvements. It also outperforms existing activation functions (regardless whether the latter are fixed or trainable) in tasks such as semantic segmentation, image generation, regression problems, and image classification. Our code is available at https://github.com/BGU-CS-VIL/DiTAC.
Accurate predictions of base stations' traffic load are essential to mobile cellular operators and their users as they support the efficient use of network resources and sustain smart cities and roads. Traditionally, cellular network time-series have been considered for this prediction task. More recently, exogenous factors such as points of presence and other environmental knowledge have been introduced to facilitate cellular traffic forecasting. In this study, we focus on smart roads and explore road traffic measures to model the processes underlying cellular traffic generation with the goal to improve prediction performance. Comprehensive experiments demonstrate that by employing road flow and speed, in addition to cellular network metrics, cellular load prediction errors can be reduced by as much as 56.5 %. The code and more detailed results are available on https://github.com/nvassileva/DataMatters.
Antenna modeling is a time-consuming and complex process, decreasing the speed of antenna analysis and design. In this paper, a large language model (LLM)- enabled antenna modeling method, called LEAM, is presented to address this challenge. LEAM enables automatic antenna model generation based on language descriptions via prompt input, images, descriptions from academic papers, patents, and technical reports (either one or multiple). The effectiveness of LEAM is demonstrated by three examples: a Vivaldi antenna generated from a complete user description, a slotted patch antenna generated from an incomplete user description and the operating frequency, and a monopole slotted antenna generated from images and descriptions scanned from the literature. For all the examples, correct antenna models are generated in a few minutes. The code can be accessed via https://github.com/TaoWu974/LEAM.
End-to-end speech large language models ((LLMs)) extend the capabilities of text-based models to directly process and generate audio tokens. However, this often leads to a decline in reasoning and generation performance compared to text input, a phenomenon referred to as intelligence degradation. To systematically evaluate this gap, we propose S2SBench, a benchmark designed to quantify performance degradation in Speech LLMs. It includes diagnostic datasets targeting sentence continuation and commonsense reasoning under audio input. We further introduce a pairwise evaluation protocol based on perplexity differences between plausible and implausible samples to measure degradation relative to text input. We apply S2SBench to analyze the training process of Baichuan-Audio, which further demonstrates the benchmark's effectiveness. All datasets and evaluation code are available at https://github.com/undobug/S2SBench.
Due to its effectivity and efficiency, deep hashing approaches are widely used for large-scale visual search. However, it is still challenging to produce compact and discriminative hash codes for images associated with multiple semantics for two main reasons, 1) similarity constraints designed in most of the existing methods are based upon an oversimplified similarity assignment(i.e., 0 for instance pairs sharing no label, 1 for instance pairs sharing at least 1 label), 2) the exploration in multi-semantic relevance are insufficient or even neglected in many of the existing methods. These problems significantly limit the discrimination of generated hash codes. In this paper, we propose a novel self-supervised asymmetric deep hashing method with a margin-scalable constraint(SADH) approach to cope with these problems. SADH implements a self-supervised network to sufficiently preserve semantic information in a semantic feature dictionary and a semantic code dictionary for the semantics of the given dataset, which efficiently and precisely guides a feature learning network to preserve multilabel semantic information using an asymmetric learning strategy. By further exploiting semantic dictionaries, a new margin-scalable constraint is employed for both precise similarity searching and robust hash code generation. Extensive empirical research on four popular benchmarks validates the proposed method and shows it outperforms several state-of-the-art approaches.
Recent advancements in large language models (LLMs) have extended their capabilities to handle long contexts. However, increasing the number of model layers and the length of input sequences significantly escalates the memory required to store key-value (KV) cache, posing challenges for efficient inference. To mitigate this issue, we present SimLayerKV, a simple yet effective method that reduces inter-layer KV cache redundancies by selectively dropping cache in identified lazy layers. Our approach is based on the observation that certain layers in long-context LLMs exhibit "lazy" behavior, contributing less to modeling long-range dependencies compared to non-lazy layers. By analyzing attention weight patterns, we find that the behavior of these lazy layers is consistent across tokens during generation for a given input. This insight motivates our SimLayerKV, which identifies lazy layers and reduces their KV cache accordingly. SimLayerKV is training-free, generalizable, and can be implemented with only seven lines of code. We conduct extensive experiments on three representative LLMs, e.g., LLaMA2-7B, LLaMA3-8B, and Mistral-7B across 16 tasks from the LongBench benchmark. The results demonstrate that SimLayerKV achieves a KV cache compression ratio of 5$\times$ with only a 1.2% performance drop when combined with 4-bit quantization. Our code is available at https://github.com/sail-sg/SimLayerKV.
LLM-based agents have shown promising capabilities in a growing range of software engineering (SWE) tasks. However, advancing this field faces two critical challenges. First, high-quality training data is scarce, especially data that reflects real-world SWE scenarios, where agents must interact with development environments, execute code and adapt behavior based on the outcomes of their actions. Existing datasets are either limited to one-shot code generation or comprise small, manually curated collections of interactive tasks, lacking both scale and diversity. Second, the lack of fresh interactive SWE tasks affects evaluation of rapidly improving models, as static benchmarks quickly become outdated due to contamination issues. To address these limitations, we introduce a novel, automated, and scalable pipeline to continuously extract real-world interactive SWE tasks from diverse GitHub repositories. Using this pipeline, we construct SWE-rebench, a public dataset comprising over 21,000 interactive Python-based SWE tasks, suitable for reinforcement learning of SWE agents at scale. Additionally, we use continuous supply of fresh tasks collected using SWE-rebench methodology to build a contamination-free benchmark for agentic software engineering. We compare results of various LLMs on this benchmark to results on SWE-bench Verified and show that performance of some language models might be inflated due to contamination issues.
Recent work by Clark et al. (2020) shows that transformers can act as 'soft theorem provers' by answering questions over explicitly provided knowledge in natural language. In our work, we take a step closer to emulating formal theorem provers, by proposing PROVER, an interpretable transformer-based model that jointly answers binary questions over rule-bases and generates the corresponding proofs. Our model learns to predict nodes and edges corresponding to proof graphs in an efficient constrained training paradigm. During inference, a valid proof, satisfying a set of global constraints is generated. We conduct experiments on synthetic, hand-authored, and human-paraphrased rule-bases to show promising results for QA and proof generation, with strong generalization performance. First, PROVER generates proofs with an accuracy of 87%, while retaining or improving performance on the QA task, compared to RuleTakers (up to 6% improvement on zero-shot evaluation). Second, when trained on questions requiring lower depths of reasoning, it generalizes significantly better to higher depths (up to 15% improvement). Third, PROVER obtains near perfect QA accuracy of 98% using only 40% of the training data. However, generating proofs for questions requiring higher depths of reasoning becomes challenging, and the accuracy drops to 65% for 'depth 5', indicating significant scope for future work. Our code and models are publicly available at https://github.com/swarnaHub/PRover
The use of computer vision for product and assembly quality control is becoming ubiquitous in the manufacturing industry. Lately, it is apparent that machine learning based solutions are outperforming classical computer vision algorithms in terms of performance and robustness. However, a main drawback is that they require sufficiently large and labeled training datasets, which are often not available or too tedious and too time consuming to acquire. This is especially true for low-volume and high-variance manufacturing. Fortunately, in this industry, CAD models of the manufactured or assembled products are available. This paper introduces CAD2Render, a GPU-accelerated synthetic data generator based on the Unity High Definition Render Pipeline (HDRP). CAD2Render is designed to add variations in a modular fashion, making it possible for high customizable data generation, tailored to the needs of the industrial use case at hand. Although CAD2Render is specifically designed for manufacturing use cases, it can be used for other domains as well. We validate CAD2Render by demonstrating state of the art performance in two industrial relevant setups. We demonstrate that the data generated by our approach can be used to train object detection and pose estimation models with a high enough accuracy to direct a robot. The code for CAD2Render is available at https://github.com/EDM-Research/CAD2Render.
Recent strides in Large Language Models (LLMs) have saturated many Natural Language Processing (NLP) benchmarks, emphasizing the need for more challenging ones to properly assess LLM capabilities. However, domain-specific and multilingual benchmarks are rare because they require in-depth expertise to develop. Still, most public models are trained predominantly on English corpora, while other languages remain understudied, particularly for practical domain-specific NLP tasks. In this work, we introduce a novel NLP benchmark for the legal domain that challenges LLMs in five key dimensions: processing \emph{long documents} (up to 50K tokens), using \emph{domain-specific knowledge} (embodied in legal texts), \emph{multilingual} understanding (covering five languages), \emph{multitasking} (comprising legal document-to-document Information Retrieval, Court View Generation, Leading Decision Summarization, Citation Extraction, and eight challenging Text Classification tasks) and \emph{reasoning} (comprising especially Court View Generation, but also the Text Classification tasks). Our benchmark contains diverse datasets from the Swiss legal system, allowing for a comprehensive study of the underlying non-English, inherently multilingual legal system. Despite the large size of our datasets (some with hundreds of thousands of examples), existing publicly available multilingual models struggle with most tasks, even after extensive in-domain pre-training and fine-tuning. We publish all resources (benchmark suite, pre-trained models, code) under permissive open CC BY-SA licenses.
In recent years, studies on image generation models of spiking neural networks (SNNs) have gained the attention of many researchers. Variational autoencoders (VAEs), as one of the most popular image generation models, have attracted a lot of work exploring their SNN implementation. Due to the constrained binary representation in SNNs, existing SNN VAE methods implicitly construct the latent space by an elaborated autoregressive network and use the network outputs as the sampling variables. However, this unspecified implicit representation of the latent space will increase the difficulty of generating high-quality images and introduces additional network parameters. In this paper, we propose an efficient spiking variational autoencoder (ESVAE) that constructs an interpretable latent space distribution and design a reparameterizable spiking sampling method. Specifically, we construct the prior and posterior of the latent space as a Poisson distribution using the firing rate of the spiking neurons. Subsequently, we propose a reparameterizable Poisson spiking sampling method, which is free from the additional network. Comprehensive experiments have been conducted, and the experimental results show that the proposed ESVAE outperforms previous SNN VAE methods in reconstructed & generated images quality. In addition, experiments demonstrate that ESVAE's encoder is able to retain the original image information more efficiently, and the decoder is more robust. The source code is available at https://github.com/QgZhan/ESVAE.
Collage is a creative art form that uses diverse material scraps as a base unit to compose a single image. Although pixel-wise generation techniques can reproduce a target image in collage style, it is not a suitable method due to the solid stroke-by-stroke nature of the collage form. While some previous works for stroke-based rendering produced decent sketches and paintings, collages have received much less attention in research despite their popularity as a style. In this paper, we propose a method for learning to make collages via reinforcement learning without the need for demonstrations or collage artwork data. We design the collage Markov Decision Process (MDP), which allows the agent to handle various materials and propose a model-based soft actor-critic to mitigate the agent's training burden derived from the sophisticated dynamics of collage. Moreover, we devise additional techniques such as active material selection and complexity-based multi-scale collage to handle target images at any size and enhance the results' aesthetics by placing relatively more scraps in areas of high complexity. Experimental results show that the trained agent appropriately selected and pasted materials to regenerate the target image into a collage and obtained a higher evaluation score on content and style than pixel-wise generation methods. Code is available at https://github.com/northadventure/CollageRL.