unsupervised-learning

This paper proposes a locality-aware private class identification approach and a reliable optimal transport-based method (ReOT) to address domain adaptation challenges under extreme label shift, particularly distinguishing shared from private classes.

This paper introduces MultiLinguahah, an unsupervised multilingual method for acoustic laughter segmentation using Isolation Forests on BYOL-A encoder representations. The authors demonstrate that their approach outperforms state-of-the-art supervised methods in non-English settings by treating laughter detection as an anomaly detection task.

MSR-MEL introduces an unsupervised framework using LLMs to synthesize and reason over multi-perspective evidence for multimodal entity linking, outperforming prior methods on standard benchmarks.

Researchers from the University of British Columbia propose an unsupervised graph-based system for organizing arguments from online debates by constructing interaction graphs and applying community detection to reveal diverse viewpoint distributions. The approach requires no training data and aims to help users navigate complex argumentative landscapes and combat filter bubbles.

C-Mining proposes an unsupervised framework for discovering cultural seeds in LLM training data by exploiting cross-lingual geometric misalignment in embedding spaces, enabling scalable synthetic data generation for cultural alignment without manual or LLM supervision.

The paper introduces diverse dictionary learning, showing that key set-theoretic relationships among latent variables can be identified from observational data without strong assumptions, enabling partial or full identifiability with minimal inductive bias.

OpenAI presents a contrastive pre-training approach for generating high-quality text and code embeddings at scale without supervision, achieving state-of-the-art results on linear-probe classification, semantic search, and code search benchmarks.

OpenAI's Image GPT (iGPT) applies GPT-2 transformers to pixel sequences for image generation and classification, demonstrating that the same architecture used for language can learn coherent visual features in an unsupervised manner and achieve competitive performance on image classification benchmarks.

OpenAI released MuseNet, a deep neural network based on GPT-2 architecture that generates 4-minute musical compositions with 10 instruments by learning patterns from hundreds of thousands of MIDI files. The model can combine multiple music styles and blend them in novel ways.

OpenAI introduces GPT-2, a 1.5 billion parameter transformer-based language model trained on 40GB of internet text that achieves state-of-the-art performance on language modeling benchmarks and demonstrates zero-shot capabilities in reading comprehension, translation, question answering, and summarization. Due to safety concerns, only a smaller model and technical paper are released publicly rather than the full trained model.

OpenAI presents a two-stage approach for improving language understanding: pretraining a transformer model on large unsupervised datasets using language modeling, then fine-tuning on smaller supervised datasets for specific tasks. The method achieves state-of-the-art results across diverse tasks including commonsense reasoning, semantic similarity, and reading comprehension with minimal hyperparameter tuning.

OpenAI demonstrates an unsupervised system that learns sentiment representation by training a multiplicative LSTM to predict the next character in Amazon reviews, achieving state-of-the-art sentiment analysis on Stanford Sentiment Treebank (91.8% accuracy) while requiring 30-100x fewer labeled examples than supervised approaches. The model discovers a distinct 'sentiment neuron' that captures sentiment information and can be directly manipulated to control text generation sentiment.

OpenAI presents a method for unsupervised third-person imitation learning that enables agents to learn from demonstrations taken from different viewpoints without explicit state correspondence, using domain confusion techniques to learn viewpoint-agnostic features.

OpenAI researchers present a Variational Lossy Autoencoder (VLAE) that combines VAEs with neural autoregressive models (RNN, MADE, PixelRNN/CNN) to learn controllable global representations, achieving state-of-the-art results on MNIST, OMNIGLOT, and Caltech-101 Silhouettes density estimation tasks.

OpenAI publishes an overview of generative models as an approach to developing machine understanding of the world, explaining how these models work by learning to generate data similar to their training sets and their potential applications across various domains.