arXiv research

New research shows LLMs can't be explained by statistical generalization alone.

Study shows LLMs can extrapolate rules from out-of-distribution prompts.

Paper tackles multi-object reinforcement learning, improving skill extrapolation.

pRSL combines probabilistic rules to improve multi-label classification.

Wittgenstein's Rule Following evolves datasets by extrapolating structural descriptors.

Neural networks struggle with TSP beyond small instances, requiring new approaches.

Convex sparsity-inducing regularizations are ubiquitous in high-dimensional machine learning, but solving the resulting optimization problems can be slow. To accelerate solvers, state-of-the-art approaches consist in reducing the size of the optimization problem at hand. In the context of regression, this can be achiev…

Adapts pretrained models to new classes without additional training.

The generalization properties of Gaussian processes depend heavily on the choice of kernel, and this choice remains a dark art. We present the Neural Kernel Network (NKN), a flexible family of kernels represented by a neural network. The NKN architecture is based on the composition rules for kernels, so that each unit …

Martingale Doppelgänger-Eval benchmarks VLMs on candlestick evidence vs. trend extrapolation

Temporal Causal Prior-Data Fitted Networks (TCPFN) for industrial time series causal discovery

Zero-shot anomaly detection method using batch normalization.

Paper proposes redundancy-free features for zero-shot object recognition.

Neural production systems learn visual dynamics by applying rule templates to entities.

Zero-shot learning transfers knowledge from seen classes to novel unseen classes to reduce human labor of labelling data for building new classifiers. Much effort on zero-shot learning however has focused on the standard multi-class setting, the more challenging multi-label zero-shot problem has received limited attent…

Study shows zero-shot super-resolution in neural operators is impossible in many cases.

RTE enables extrapolation to new tasks by learning task transformations.

The ability to explain decisions made by AI systems is highly sought after, especially in domains where human lives are at stake such as medicine or autonomous vehicles. While it is often possible to approximate the input-output relations of deep neural networks with a few human-understandable rules, the discovery of t…

We extend nonparametric models to handle extrapolation, providing bounds for inference.

Aggregates diverse zero-shot LLM outputs for better corporate disclosure classification.

Generalization and reliability of multilingual translation often highly depend on the amount of available parallel data for each language pair of interest. In this paper, we focus on zero-shot generalization---a challenging setup that tests models on translation directions they have not been optimized for at training t…

Paper improves zero-shot protein stability prediction by clarifying free-energy foundations.

This study evaluates zero-shot LLMs in finance, finding ChatGPT performs well but fine-tuned models are better.

Study uses zero-shot models to forecast mortality rates globally.

As modern deep networks become more complex, and get closer to human-like capabilities in certain domains, the question arises of how the representations and decision rules they learn compare to the ones in humans. In this work, we study representations of sentences in one such artificial system for natural language pr…

We address the problem of learning fine-grained cross-modal representations. We propose an instance-based deep metric learning approach in joint visual and textual space. The key novelty of this paper is that it shows that using per-image semantic supervision leads to substantial improvement in zero-shot performance ov…

Binary encoding enables neural networks to extrapolate periodic functions.

ZegOT uses optimal transport to zero-shot segment images with text prompts.

Because it is not feasible to collect training data for every language, there is a growing interest in cross-lingual transfer learning. In this paper, we systematically explore zero-shot cross-lingual transfer learning on reading comprehension tasks with a language representation model pre-trained on multi-lingual corp…

Method controls extrapolation in prediction profiles for statistical and machine learning models.

Neural networks struggle with extrapolation, but a new framework allows them to learn counterfactual invariances.

Paper proposes CCVAE for generalized zero-shot domain adaptation.

Dissertation tackles zero-shot anomaly detection, focusing on consistent anomalies and proposing CoDeGraph framework.

New method optimizes language model performance for test-time strategies.

This work improves fairness in federated learning by using zero-shot data augmentation.

Theory explains how AI models can predict unseen tasks without labeled data.

M2M tackles zero-shot structured noise suppression in images.

CLIP learns joint image-text representations for zero-shot learning.

We introduce a new weight-decay scaling rule to maintain sublayer gains across different widths in modern scale-invariant architectures.

Zero-shot understanding of accidents from surveillance videos using vision-language models

New methods reduce extrapolation errors in feature importance.

With the recent renaissance of deep convolution neural networks, encouraging breakthroughs have been achieved on the supervised recognition tasks, where each class has sufficient training data and fully annotated training data. However, to scale the recognition to a large number of classes with few or now training samp…

New models extrapolate false alarms in ASV without new data.

Neural networks extrapolate poorly in simple tasks but succeed in complex ones.

A new principle for extrapolating regression outside training data.

Zero-shot contrastive loss improves text-guided image style transfer without extra training.

Faster ZSL with continual learning and self-gating.

We extend return extrapolation to incorporate asymmetry and saturation, finding that asymmetric nonlinear extrapolation leads to lower welfare loss.