Adapting to the Unknown: Training-Free Audio-Visual Event Perception with Dynamic Thresholds

In the domain of audio-visual event perception, which focuses on the temporal localization and classification of events across distinct modalities (audio and visual), existing approaches are constrained by the vocabulary available in their training data. This limitation significantly impedes their capacity to generalize to novel, unseen event categories. Furthermore, the annotation process for this task is labor-intensive, requiring extensive manual labeling across modalities and temporal segments, limiting the scalability of current methods. Current state-of-the-art models ignore the shifts in event distributions over time, reducing their ability to adjust to changing video dynamics. Additionally, previous methods rely on late fusion to combine audio and visual information. While straightforward, this approach results in a significant loss of multimodal interactions. To address these challenges, we propose Audio-Visual Adaptive Video Analysis ($\text{AV}^2\text{A}$), a model-agnostic approach that requires no further training and integrates a score-level fusion technique to retain richer multimodal interactions. $\text{AV}^2\text{A}$ also includes a within-video label shift algorithm, leveraging input video data and predictions from prior frames to dynamically adjust event distributions for subsequent frames. Moreover, we present the first training-free, open-vocabulary baseline for audio-visual event perception, demonstrating that $\text{AV}^2\text{A}$ achieves substantial improvements over naive training-free baselines. We demonstrate the effectiveness of $\text{AV}^2\text{A}$ on both zero-shot and weakly-supervised state-of-the-art methods, achieving notable improvements in performance metrics over existing approaches.

VideoJAM: Joint Appearance-Motion Representations for Enhanced Motion Generation in Video Models

Despite tremendous recent progress, generative video models still struggle to capture real-world motion, dynamics, and physics. We show that this limitation arises from the conventional pixel reconstruction objective, which biases models toward appearance fidelity at the expense of motion coherence. To address this, we introduce VideoJAM, a novel framework that instills an effective motion prior to video generators, by encouraging the model to learn a joint appearance-motion representation. VideoJAM is composed of two complementary units. During training, we extend the objective to predict both the generated pixels and their corresponding motion from a single learned representation. During inference, we introduce Inner-Guidance, a mechanism that steers the generation toward coherent motion by leveraging the model's own evolving motion prediction as a dynamic guidance signal. Notably, our framework can be applied to any video model with minimal adaptations, requiring no modifications to the training data or scaling of the model. VideoJAM achieves state-of-the-art performance in motion coherence, surpassing highly competitive proprietary models while also enhancing the perceived visual quality of the generations. These findings emphasize that appearance and motion can be complementary and, when effectively integrated, enhance both the visual quality and the coherence of video generation. Project website: https://hila-chefer.github.io/videojam-paper.github.io/

On the Expressivity of Selective State-Space Layers: A Multivariate Polynomial Approach

Recent advances in efficient sequence modeling have introduced selective state-space layers, a key component of the Mamba architecture, which have demonstrated remarkable success in a wide range of NLP and vision tasks. While Mamba's empirical performance has matched or surpassed SoTA transformers on such diverse benchmarks, the theoretical foundations underlying its powerful representational capabilities remain less explored. In this work, we investigate the expressivity of selective state-space layers using multivariate polynomials, and prove that they surpass linear transformers in expressiveness. Consequently, our findings reveal that Mamba offers superior representational power over linear attention-based models for long sequences, while not sacrificing their generalization. Our theoretical insights are validated by a comprehensive set of empirical experiments on various datasets.

Deep Active Speech Cancellation with Multi-Band Mamba Network

We present a novel deep learning network for Active Speech Cancellation (ASC), advancing beyond Active Noise Cancellation (ANC) methods by effectively canceling both noise and speech signals. The proposed Multi-Band Mamba architecture segments input audio into distinct frequency bands, enabling precise anti-signal generation and improved phase alignment across frequencies. Additionally, we introduce an optimization-driven loss function that provides near-optimal supervisory signals for anti-signal generation. Experimental results demonstrate substantial performance gains, achieving up to 7.2dB improvement in ANC scenarios and 6.2dB in ASC, significantly outperforming existing methods. Audio samples are available at https://mishalydev.github.io/DeepASC-Demo

Classifier-Guided Captioning Across Modalities

Most current captioning systems use language models trained on data from specific settings, such as image-based captioning via Amazon Mechanical Turk, limiting their ability to generalize to other modality distributions and contexts. This limitation hinders performance in tasks like audio or video captioning, where different semantic cues are needed. Addressing this challenge is crucial for creating more adaptable and versatile captioning frameworks applicable across diverse real-world contexts. In this work, we introduce a method to adapt captioning networks to the semantics of alternative settings, such as capturing audibility in audio captioning, where it is crucial to describe sounds and their sources. Our framework consists of two main components: (i) a frozen captioning system incorporating a language model (LM), and (ii) a text classifier that guides the captioning system. The classifier is trained on a dataset automatically generated by GPT-4, using tailored prompts specifically designed to enhance key aspects of the generated captions. Importantly, the framework operates solely during inference, eliminating the need for further training of the underlying captioning model. We evaluate the framework on various models and modalities, with a focus on audio captioning, and report promising results. Notably, when combined with an existing zero-shot audio captioning system, our framework improves its quality and sets state-of-the-art performance in zero-shot audio captioning.

Segment-Based Attention Masking for GPTs

Modern Language Models (LMs) owe much of their success to masked causal attention, the backbone of Generative Pre-Trained Transformer (GPT) models. Although GPTs can process the entire user prompt at once, the causal masking is applied to all input tokens step-by-step, mimicking the generation process. This imposes an unnecessary constraint during the initial "prefill" phase when the model processes the input prompt and generates the internal representations before producing any output tokens. In this work, attention is masked based on the known block structure at the prefill phase, followed by the conventional token-by-token autoregressive process after that. For example, in a typical chat prompt, the system prompt is treated as one block, and the user prompt as the next one. Each of these is treated as a unit for the purpose of masking, such that the first tokens in each block can access the subsequent tokens in a non-causal manner. Then, the model answer is generated in the conventional causal manner. This Segment-by-Segment scheme entails no additional computational overhead. When integrating it into models such as Llama and Qwen, state-of-the-art performance is consistently achieved.

Reversed Attention: On The Gradient Descent Of Attention Layers In GPT

The success of Transformer-based Language Models (LMs) stems from their attention mechanism. While this mechanism has been extensively studied in explainability research, particularly through the attention values obtained during the forward pass of LMs, the backward pass of attention has been largely overlooked. In this work, we study the mathematics of the backward pass of attention, revealing that it implicitly calculates an attention matrix we refer to as "Reversed Attention". We examine the properties of Reversed Attention and demonstrate its ability to elucidate the models' behavior and edit dynamics. In an experimental setup, we showcase the ability of Reversed Attention to directly alter the forward pass of attention, without modifying the model's weights, using a novel method called "attention patching". In addition to enhancing the comprehension of how LM configure attention layers during backpropagation, Reversed Attention maps contribute to a more interpretable backward pass.

Diffusion-Based Attention Warping for Consistent 3D Scene Editing

We present a novel method for 3D scene editing using diffusion models, designed to ensure view consistency and realism across perspectives. Our approach leverages attention features extracted from a single reference image to define the intended edits. These features are warped across multiple views by aligning them with scene geometry derived from Gaussian splatting depth estimates. Injecting these warped features into other viewpoints enables coherent propagation of edits, achieving high fidelity and spatial alignment in 3D space. Extensive evaluations demonstrate the effectiveness of our method in generating versatile edits of 3D scenes, significantly advancing the capabilities of scene manipulation compared to the existing methods. Project page: \url{https://attention-warp.github.io}

SphereUFormer: A U-Shaped Transformer for Spherical 360 Perception

This paper proposes a novel method for omnidirectional 360$\degree$ perception. Most common previous methods relied on equirectangular projection. This representation is easily applicable to 2D operation layers but introduces distortions into the image. Other methods attempted to remove the distortions by maintaining a sphere representation but relied on complicated convolution kernels that failed to show competitive results. In this work, we introduce a transformer-based architecture that, by incorporating a novel ``Spherical Local Self-Attention'' and other spherically-oriented modules, successfully operates in the spherical domain and outperforms the state-of-the-art in 360$\degree$ perception benchmarks for depth estimation and semantic segmentation.

Add-it: Training-Free Object Insertion in Images With Pretrained Diffusion Models

Adding Object into images based on text instructions is a challenging task in semantic image editing, requiring a balance between preserving the original scene and seamlessly integrating the new object in a fitting location. Despite extensive efforts, existing models often struggle with this balance, particularly with finding a natural location for adding an object in complex scenes. We introduce Add-it, a training-free approach that extends diffusion models' attention mechanisms to incorporate information from three key sources: the scene image, the text prompt, and the generated image itself. Our weighted extended-attention mechanism maintains structural consistency and fine details while ensuring natural object placement. Without task-specific fine-tuning, Add-it achieves state-of-the-art results on both real and generated image insertion benchmarks, including our newly constructed "Additing Affordance Benchmark" for evaluating object placement plausibility, outperforming supervised methods. Human evaluations show that Add-it is preferred in over 80% of cases, and it also demonstrates improvements in various automated metrics.