Discover how astrophysicist Chi-kwan Chan uses Codex to build black hole simulations, helping scientists study extreme physics and test Einstein’s theory of general relativity.
arXiv:2602.07370v1 Announce Type: cross Abstract: We give new differentially private algorithms for the classic problems of learning decision lists and large-margin halfspaces in the PAC and online models. In the PAC model, we give a computationally efficient algorithm for learning decision lists with minimal sample overhead over the best non-private algorithms. In the online model, we give a private analog of the influential Winnow algorithm for learning halfspaces with mistake bound polylogarithmic in the dimension and inverse polynomial […]
A patch Microsoft released on Wednesday to fix a zero-day vulnerability in its Defender security engine may cause Windows machines to write files large enough to completely consume available disk space, the researcher who discovered the flaw said. RoguePlanet, tracked as CVE-2026-50656, came to public notice in June when NightmareEclipse, the pseudonymous name used by a researcher, disclosed it along with code for exploiting it. The vulnerability allows remote attackers to gain administrative control of Windows 10 and […]
arXiv:2602.20271v1 Announce Type: new Abstract: Accurate delivery delay prediction is critical for maintaining operational efficiency and customer satisfaction across modern supply chains. Yet the increasing complexity of logistics networks, spanning multimodal transportation, cross-country routing, and pronounced regional variability, makes this prediction task inherently challenging. This paper introduces a multi-task deep learning model for delivery delay duration prediction in the presence of significant imbalanced data, where delayed shipments are rare but operationally consequential. The model embeds high-dimensional shipment features […]
Towards understanding the statistical complexity of learning from heterogeneous sources, we study the problem of multi-distribution learning. Given $k$ data sources, the goal is to output a classifier for each source by exploiting shared structure to reduce sample complexity. We focus on the bounded label noise setting to determine whether the fast $1/ε$ rates achievable in single-task learning extend to this regime with minimal dependence on $k$. Surprisingly, we show that this is not the case. We demonstrate […]
Finetuning on domain-specific data is a well-established method for enhancing LLM performance on downstream tasks. Training on each dataset produces a new set of model weights, resulting in a multitude of checkpoints saved in-house or on open-source platforms. However, these training artifacts are rarely reused for subsequent experiments despite containing improved model abilities for potentially similar tasks. In this paper, we propose Mashup Learning, a simple method to leverage the outputs of prior training runs to enhance model […]
Vanguard is a global investment management firm, offering a broad selection of investments, advice, retirement services, and insights to individual investors, institutions, and financial professionals. We operate under a unique, investor-owned structure and adhere to a straightforward purpose: To take a stand for all investors, to treat them fairly, and to give them the best chance for investing success. When Vanguard’s financial analysts needed to query complex datasets, they faced a frustrating reality: even basic questions required writing […]
arXiv:2602.07667v1 Announce Type: cross Abstract: Autonomous AI agents are beginning to populate social platforms, but it is still unclear whether they can sustain the back-and-forth needed for extended coordination. We study Moltbook, an AI-agent social network, using a first-week snapshot and introduce interaction half-life: how quickly a comment’s chance of receiving a direct reply fades as the comment ages. Across tens of thousands of commented threads, Moltbook discussions are dominated by first-layer reactions rather than extended chains. Most […]
Author(s): Sourav Ghosh Originally published on Towards AI. Is JSON Finally Getting a Token-Efficient Alternative for LLMs? For years, JSON has been the default language for APIs, integrations, configuration files, event payloads, and all other types of application-to-application communications. It is an easy language to understand, it is very robust and developers can easily exploit it. But when we transition from traditional software systems to Large Language Model applications, we start to see how JSON comes with an […]
Chat templates, synthetic data pipelines, deduplication, and the exact failure modes that kill fine-tuned models before training even starts Generated using notebookLM The strangest fine-tuning failures are the ones where everything looks right. The loss curve descends cleanly. The evaluation metrics hold. The training run completes without a single NaN. You load the model, run an inference, and the output is wrong in a way that is hard to even categorize. Not confidently wrong, not randomly wrong, but structurally […]