unsupervised-learning

Presents Oat, a lightweight unsupervised method for identifying error steps in LLM-based agentic failure trajectories using neural controlled differential equations trained only on successful trajectories. It achieves 200-5000x speedup over prompting baselines with significant F1 improvements in in-domain and out-of-distribution settings.

This paper proposes C2TSP, an end-to-end unsupervised learning method for the Traveling Salesman Problem that learns a tractable distribution over near-tour structures using a connected-by-construction Gibbs family, incorporating implicit differentiation and certificate-guided sharpening to preserve interpretable Hamiltonian structure.

This paper demonstrates that unsupervised visual pretraining on documents, without text extraction, consistently outperforms text-only pretraining for language intelligence, providing an efficient and scalable approach for foundation models.

Proposes a novel self-supervised image clustering framework that uses an evolution-strategy outer loop to maximize a 'surprise score' without needing a per-step loss, paired with a gradient-descent inner loop, achieving state-of-the-art results on standard benchmarks in the strict non-parametric setting.

Proposes TENSOR, an unsupervised anomaly detection method for identifying information operations users by analyzing temporal behavioral and language patterns using temporal point processes and LLM responses. Outperforms baselines on five real-world datasets.

This paper proposes SCISE, a scalable unsupervised graph clustering framework that uses community-aware sampling and structural entropy to overcome structural isolation in mini-batch training, achieving state-of-the-art results on benchmark datasets.

Proposes a novel distance metric for intra-attribute distances in categorical data clustering that treats nominal and ordinal attributes separately while preserving order relationships, and introduces a clustering algorithm that integrates weight learning and partitioning.

This paper presents an unsupervised clustering-based framework using K-Means++ to detect suspicious trading patterns in capital market data, achieving a silhouette score of 0.561 and identifying 2.02% of trades as potentially fraudulent.

Proposes a novel framework combining active learning with masked reconstruction and minimax strategies to improve unsupervised time series anomaly detection, achieving 12.39% AUC improvement over baselines across 28 test cases.

NEO is a new type of world model that learns to discover reusable building blocks of explanation from raw observations without supervision or language, selected as an ICML 2026 oral presentation.

RGLD combines global random-feature density estimation with local neighbor analysis for efficient unsupervised tabular anomaly detection, achieving top AUROC performance on 47 datasets while being 50x-580x faster than deep detectors.

AnTenA is an interpretable system that leverages large language models to generate natural language explanations for hidden patterns extracted via tensor decomposition, without relying on potentially inaccurate labels or metadata.

Proposes a Wasserstein-GAN approach for unsupervised calibration of sensor-induced distribution drifts, validated on tracking detector toy models and simulated calorimeter data with aging effects.

MM++ is a fully unsupervised, post-hoc framework for out-of-distribution detection that fuses discriminative intermediate layers via top-K gated feature fusion and uses a regularized tied covariance matrix for scale-invariant distance estimation.

Introduces Temporal Difference in Vision (TDV), a novel visual representation learning paradigm that learns useful representations without augmentations, masking, cropping, or reconstruction, and matches state-of-the-art methods on dense spatial tasks.

DataCOPE is an unsupervised verifier-guided skill discovery framework for data-analytic agents that derives verifier signals from exploration trajectories without labeled supervision. It improves performance by 9.71% and 32.30% on report-style and reasoning-style data analysis tasks respectively.

This paper introduces coherence, a geometric constraint for neural representations inspired by grid cells and head direction cells in the brain. Coherence ensures that features respond to geometrically connected regions of the data manifold, improving interpretability; the authors propose a differentiable objective (Coh) and validate it on synthetic data, rotated MNIST, and BERT token embeddings.

This paper introduces Internal Coherence Maximization (ICM) to generate persona-specific examples for aligning AI with diverse human values without human supervision, demonstrating that coherent examples generalize better across benchmarks.

This paper introduces MADQI, a composite metric for evaluating unsupervised anomaly detection in maritime AIS data without requiring labeled data. The framework combines four metrics and achieves an 80.37% MADQI score on test datasets.

This paper presents a single-stage sparse coding method using unsupervised sparse autoencoders and natural inverted indexing to accelerate multi-vector retrieval, outperforming traditional k-means based approaches.