arXiv:2602.08782v1 Announce Type: new Abstract: One of the core facets of Bayesianism is in the updating of prior beliefs in light of new evidence$text{ — }$so how can we maintain a Bayesian approach if we have no prior beliefs in the first place? This is one of the central challenges in the field of Bayesian deep learning, where it is not clear how to represent beliefs about a prediction task by prior distributions over model parameters. Bridging the […]

Ver mais

:::info Authors: Dani Valevski Yaniv Leviathan Moab Arar Shlomi Fruchter ::: Table of Links ABSTRACT 1 INTRODUCTION 2 INTERACTIVE WORLD SIMULATION 3 GAMENGEN 3.1 DATA COLLECTION VIA AGENT PLAY 3.2 TRAINING THE GENERATIVE DIFFUSION MODEL 4 EXPERIMENTAL SETUP 4.1 AGENT TRAINING 4.2 GENERATIVE MODEL TRAINING 5 RESULTS 5.1 SIMULATION QUALITY 5.2 ABLATIONS 6 RELATED WORK 7 DISCUSSION, ACKNOWLEDGEMENTS AND REFERENCES ABSTRACT We present GameNGen, the first game engine powered entirely by a neural model that also enables real-time […]

Ver mais

Welcome to ALT Highlights, a series of blog posts spotlighting various happenings at the recent conference ALT 2021, including plenary talks, tutorials, trends in learning theory, and more! To reach a broad audience, the series will be disseminated as guest posts on different blogs in machine learning and theoretical computer science. John has been kind enough to host the first post in the series. This initiative is organized by the Learning Theory Alliance, and overseen by Gautam Kamath. […]

Ver mais

For two decades, product design has revolved around a stable premise: users initiate, software responds. Even as AI crept into products such as recommendation engines, predictive text, and fraud detection, the interface still framed intelligence as reactive. AI agents break that contract. An agent does not wait for instruction. It monitors context, forms intentions, takes actions, and adapts its strategy over time. It delegates to APIs, coordinates across systems, and sometimes executes decisions without asking first. In other […]

Ver mais

arXiv:2509.11379v2 Announce Type: replace Abstract: We demonstrate that learning procedures that rely on aggregated labels, e.g., label information distilled from noisy responses, enjoy robustness properties impossible without data cleaning. This robustness appears in several ways. In the context of risk consistency — when one takes the standard approach in machine learning of minimizing a surrogate (typically convex) loss in place of a desired task loss (such as the zero-one mis-classification error) — procedures using label aggregation obtain stronger […]

Ver mais

arXiv:2602.05042v1 Announce Type: new Abstract: Machine Learning (ML)-enabled systems challenge traditional Requirements Engineering (RE) and agile management due to data dependence, experimentation, and uncertain model behavior. Existing RE and agile practices remain poorly integrated and insufficiently tailored to these characteristics. This paper reports on the practical experience of applying RefineML, a requirements-focused approach for the continuous and agile refinement of ML-enabled systems, which integrates ML-tailored specification and agile management approaches with best practices derived from a systematic mapping […]

Ver mais

There is a quiet shift happening in software development. For decades, the most valuable skill for a developer was the ability to write code efficiently and correctly. Entire careers were built around mastering frameworks, memorizing patterns, and becoming faster at turning requirements into working software. But the rise of AI coding agents is changing that equation. Today, machines are becoming exceptionally good at the one thing developers spent years mastering: churning out code. Tools like OpenAI’s ChatGPT and […]

Ver mais

arXiv:2601.12304v1 Announce Type: new Abstract: Vision-language pre-training (VLP) models are vulnerable to adversarial examples, particularly in black-box scenarios. Existing multimodal attacks often suffer from limited perturbation diversity and unstable multi-stage pipelines. To address these challenges, we propose 2S-GDA, a two-stage globally-diverse attack framework. The proposed method first introduces textual perturbations through a globally-diverse strategy by combining candidate text expansion with globally-aware replacement. To enhance visual diversity, image-level perturbations are generated using multi-scale resizing and block-shuffle rotation. Extensive experiments […]

Ver mais

Large language models (LLMs) are increasingly trained in complex Reinforcement Learning, multi-agent environments, making it difficult to understand how behavior changes over training. Sparse Autoencoders (SAEs) have recently shown to be useful for data-centric interpretability. In this work, we analyze large-scale reinforcement learning training runs from the sophisticated environment of Full-Press Diplomacy by applying pretrained SAEs, alongside LLM-summarizer methods. We introduce Meta-Autointerp, a method for grouping SAE features into interpretable hypotheses about training dynamics. We discover fine-grained behaviors […]

Ver mais