Human action recognition is a fundamental and well studied problem in computer vision.

System identification has greatly benefited from deep learning techniques, particularly for modeling complex, nonlinear dynamical systems with partially unknown physics where traditional approaches may not be feasible. However, deep learning models often require large datasets and significant computational resources at training and inference due to their high-dimensional parameterizations. To address this challenge, we propose a meta-learning framework that discovers a low-dimensional manifold within the parameter space of an over-parameterized neural network architecture. This manifold is learned from a meta-dataset of input-output sequences generated by a class of related dynamical systems, enabling efficient model training while preserving the network's expressive power for the considered system class. Unlike bilevel meta-learning approaches, our method employs an auxiliary neural network to map datasets directly onto the learned manifold, eliminating the need for costly second-order gradient computations during meta-training and reducing the number of first-order updates required in inference, which could be expensive for large models. We validate our approach on a family of Bouc-Wen oscillators, which is a well-studied nonlinear system identification benchmark. We demonstrate that we are able to learn accurate models even in small-data scenarios.

We introduce the State Stream Transformer (SST), a novel LLM architecture that reveals emergent reasoning behaviours and capabilities latent in pretrained weights through addressing a fundamental limitation in traditional transformer models: the lack of latent computational continuity across autoregressive generations in the state space. SST introduces a sliding window latent state (FFN) cache with weighted decay that maintains and evolves persistent latent processes throughout autoregressive generations. Through controlled experiments comparing base and SST architectures using the same frozen weights, we demonstrate that this architectural modification alone enables enhanced reasoning capabilities which appear best explained by some form of potential higher-order processing, as evidenced by emergent metacognitive behaviours. These behaviours persist under controlled conditions designed to eliminate confounding factors such as stochastic variation or learned response patterns. Analysis of latent state distributions and processing dynamics provides evidence that it is solely the 'state stream' that is responsible for these phenomena. In quantitative evaluations, the SST achieves substantial performance improvements over the base model on two reasoning benchmarks, reaching 89.01\% accuracy on GSM-8K (0-shot) and 91.04\% on ARC Challenge (0-shot CoT). These findings indicate that persistent computation in the latent state space enables fundamentally different information processing and internal reasoning strategies, with implications for our understanding of artificial intelligence systems.

Neural Information Processing Systems http://nips.cc/

In recent years, work has gone into developing deep interpretable methods for image classification that clearly attributes a model's output to specific features of the data. One such of these methods is the Prototypical Part Network (ProtoPNet), which attempts to classify images based on meaningful parts of the input. While this method results in interpretable classifications, it often learns to classify from spurious or inconsistent parts of the image. Hoping to remedy this, we take inspiration from the recent developments in Reinforcement Learning with Human Feedback (RLHF) to fine-tune these prototypes. By collecting human annotations of prototypes quality via a 1-5 scale on the CUB-200-2011 dataset, we construct a reward model that learns human preferences and identify non-spurious prototypes. In place of a full RL update, we propose the Reweighed, Reselected, and Retrained Prototypical Part Network (R3-ProtoPNet), which adds an additional three steps to the ProtoPNet training loop. The first two steps are reward-based reweighting and reselection, which align prototypes with human feedback. The final step is retraining to realign the model's features with the updated prototypes. We find that R3-ProtoPNet improves the overall meaningfulness of the prototypes, and maintains or improves individual model performance. When multiple trained R3-ProtoPNets are incorporated into an ensemble, we find increases in both interpretability and predictive performance.

Audio-Visual Event Localization (AVEL) is the task of temporally localizing and classifying \emph{audio-visual events}, i.e., events simultaneously visible and audible in a video. In this paper, we solve AVEL in a weakly-supervised setting, where only video-level event labels (their presence/absence, but not their locations in time) are available as supervision for training. Our idea is to use a base model to estimate labels on the training data at a finer temporal resolution than at the video level and re-train the model with these labels. I.e., we determine the subset of labels for each \emph{slice} of frames in a training video by (i) replacing the frames outside the slice with those from a second video having no overlap in video-level labels, and (ii) feeding this synthetic video into the base model to extract labels for just the slice in question. To handle the out-of-distribution nature of our synthetic videos, we propose an auxiliary objective for the base model that induces more reliable predictions of the localized event labels as desired. Our three-stage pipeline outperforms several existing AVEL methods with no architectural changes and improves performance on a related weakly-supervised task as well.

Self-supervised learning (SSL) has led to great strides in speech processing. However, the resources needed to train these models has become prohibitively large as they continue to scale. Currently, only a few groups with substantial resources are capable of creating SSL models, which harms reproducibility. In this work, we optimize HuBERT SSL to fit in academic constraints. We reproduce HuBERT independently from the original implementation, with no performance loss. Our code and training optimizations make SSL feasible with only 8 GPUs, instead of the 32 used in the original work. We also explore a semi-supervised route, using an ASR model to skip the first pre-training iteration. Within one iteration of pre-training, our models improve over HuBERT on several tasks. Furthermore, our HuBERT Large variant requires only 8 GPUs, achieving similar performance to the original trained on 128. As our contribution to the community, all models, configurations, and code are made open-source in ESPnet.

An important milestone for AI is the development of algorithms that can produce drawings that are indistinguishable from those of humans. Here, we adapt the 'diversity vs. recognizability' scoring framework from Boutin et al, 2022 and find that one-shot diffusion models have indeed started to close the gap between humans and machines. However, using a finer-grained measure of the originality of individual samples, we show that strengthening the guidance of diffusion models helps improve the humanness of their drawings, but they still fall short of approximating the originality and recognizability of human drawings. Comparing human category diagnostic features, collected through an online psychophysics experiment, against those derived from diffusion models reveals that humans rely on fewer and more localized features. Overall, our study suggests that diffusion models have significantly helped improve the quality of machine-generated drawings; however, a gap between humans and machines remains -- in part explainable by discrepancies in visual strategies.