#generalization

10 curated events

papersTODAY 04:00 UTC

Paper asks whether reasoning models show systematicity on rule induction tasks

A new arXiv preprint examines whether AI reasoning models generalize systematically, meaning that learning one concept should transfer to closely related variations of it. The authors evaluate these models on rule induction tasks to test how robustly that kind of structured generalization holds. The work questions whether current reasoning capabilities reflect genuine systematic understanding or narrower pattern matching.

research/reasoning/evaluation/generalization

papersTODAY 04:00 UTC

Robusto-2 Benchmark Tests Vision-Language Models for Self-Driving in Lima and New York

A new arXiv paper introduces Robusto-2, a benchmark evaluating both humans and vision-language models on autonomous driving tasks in Lima, Peru and New York City. The work targets how well multi-modal systems generalize when deployed in unfamiliar, out-of-distribution urban environments. It is a cross-listed replacement submission on arXiv cs.AI.

benchmark/autonomous-driving/vision-language-models/generalization

papersTODAY 04:00 UTC

Paper separates task performance from compositional feature learning

A new arXiv preprint argues that strong benchmark performance does not by itself show that a model has learned compositional, environment-invariant features. The authors aim to disentangle measured accuracy from the underlying representations that support out-of-distribution generalisation, a capability often treated as a marker of biological intelligence. Their analysis is framed around how systems can transfer invariant properties from training mappings to novel compositions.

research/generalization/representation-learning/arxiv

papersTODAY 04:00 UTC

Study Finds Tabular Foundation Models Can Generalize From a Single Table

A new arXiv paper argues that deep tabular models using in-context learning can achieve broader generalization even when trained or conditioned on just one table. This challenges the common assumption that such models need many diverse tables to generalize well. The work focuses on inference-time context rather than weight updates.

tabular-data/in-context-learning/generalization/research

papersTODAY 04:00 UTC

Study Finds Structured Features Overfit Where Random Gaussian Features Grok

New research contrasts how over-parameterized ridge regression behaves with structured versus unstructured random feature maps. Prior work by Xu, Vardi and Safran proved that random Gaussian features grok, with the gap between memorization and generalization widening as the weight decay parameter shrinks. The authors report that structured features instead overfit in the regime where random features exhibit grokking.

grokking/generalization/ridge-regression/feature-learning

papersSEP 10 04:00 UTC

ROTATE: Regret-driven Open-ended Training for Ad Hoc Teamwork

A new arXiv paper presents ROTATE, a training method designed to help agents cooperate effectively with partners they have never seen before, a problem known as ad hoc teamwork. Rather than relying on a pre-built, fixed set of teammate agents followed by a separate coordination stage, the approach uses regret signals to continuously steer the creation of training collaborators. The work addresses generalization in multi-agent reinforcement learning.

multi-agent/reinforcement-learning/generalization/open-ended

papersSEP 10 04:00 UTC

Study connects f-divergence regularization and SAM through local curvature analysis

A new arXiv paper examines how divergence-based regularization relates to Sharpness-Aware Minimization, two widely used techniques for improving generalization in deep learning. The authors use local curvature analysis to explain why f-divergence regularization, like SAM, improves robustness to parameter perturbations.

deep-learning/generalization/research/optimization

papersSEP 10 04:00 UTC

CoGe-GCD paper reframes generalized category discovery with compositional generalization

A newly announced arXiv paper presents CoGe-GCD, an approach to generalized category discovery, the task of sorting unlabeled data into both known and previously unseen classes. The work draws on compositional generalization, aiming to reuse primitives learned from labeled classes while detecting when novel combinations of those primitives point to new categories. It positions GCD as a challenge requiring human-like compositional reasoning in machine learning systems.

machine-learning/clustering/generalization/unsupervised

papersSEP 10 04:00 UTC

Survey paper reviews overparameterized machine learning and the bias-variance tradeoff

A new overview article on arXiv surveys the theory of overparameterized machine learning, in which models with far more parameters than training examples still achieve strong performance. The paper explains how such behavior conflicts with the classical bias-variance tradeoff and organizes recent theoretical work developed to explain it. It serves as a structured introduction for readers interested in the statistical foundations of modern deep learning.

deep-learning/theory/generalization/survey

papersSEP 11 04:00 UTC

New Framework Quantifies Covariate and Concept Shifts in ML Generalization

A new arXiv paper proposes a general approach to measuring how covariate and concept shifts affect machine learning generalization. The authors argue that existing learning bound theory covers only narrow, idealized settings and cannot be estimated from data. Their framework aims to make distribution shift analysis broadly applicable and computable from samples.

research/distribution-shift/generalization/theory