arXiv:2602.11161v1 Announce Type: new Abstract: The web’s information ecosystem demands fact-checking systems that are both scalable and epistemically trustworthy. Automated approaches offer efficiency but often lack transparency, while human verification remains slow and inconsistent. We introduce Althea, a retrieval-augmented system that integrates question generation, evidence retrieval, and structured reasoning to support user-driven evaluation of online claims. On the AVeriTeC benchmark, Althea achieves a Macro-F1 of 0.44, outperforming standard verification pipelines and improving discrimination between supported and refuted claims. […]

Ver mais

arXiv:2603.22456v1 Announce Type: new Abstract: We describe the algorithms used by the ETH Flippers team in the CG:SHOP 2026 Challenge. Each instance consists of a set of triangulations on a common point set, and the objective is to find a central triangulation that minimizes the total parallel flip distance to the input set. Our strategy combines an exact solver for small and medium-sized instances with a suite of heuristics for larger instances. For the exact approach, we formulate […]

Ver mais

The deployment of secure design patterns in enterprise applications is a critical strategy for mitigating common vulnerabilities while enforcing a comprehensive defence-in-depth security framework. These patterns provide a reusable, systematic approach to incorporating security into software architecture, addressing typical weaknesses such as injection flaws, authentication bypasses, and data exposure. By adopting modular architectural principles, enterprises can create layered, isolated security boundaries that limit the impact of potential breaches and simplify maintenance and scalability. This paper explores the role […]

Ver mais

I used to treat Test-Driven Development (TDD) as busywork. Write a failing test. Watch it fail. Write code to pass it. Rinse and repeat. It always felt like it slowed me down. Then AI changed everything. Tools like GitHub Copilot can now generate implementation and comprehensive tests in seconds. Suddenly, the mechanical ritual of TDD seemed… optional. But the more I experimented, the more I realized something important: TDD’s real value was never the mechanics—it was the thinking […]

Ver mais

arXiv:2601.04277v1 Announce Type: new Abstract: Post-training improves large language models (LLMs) but often worsens confidence calibration, leading to systematic overconfidence. Recent unsupervised post-hoc methods for post-trained LMs (PoLMs) mitigate this by aligning PoLM confidence to that of well-calibrated pre-trained counterparts. However, framing calibration as static output-distribution matching overlooks the inference-time dynamics introduced by post-training. In particular, we show that calibration errors arise from two regimes: (i) confidence drift, where final confidence inflates despite largely consistent intermediate decision processes, […]

Ver mais

Pepeto just hit another milestone today, with the presale now above $7.43 million raised and all three ecosystem products entering final stages of completion while the rest of the crypto market is still processing the damage from the worst crash in months. Solana dropped thirty one percent in February alone and the Solana price prediction for March looks increasingly fragile, with technical patterns pointing toward further downside if key support levels fail to hold. But in the middle […]

Ver mais

arXiv:2601.10923v1 Announce Type: new Abstract: Retrieval-augmented generation (RAG) systems put more and more emphasis on grounding their responses in user-generated content found on the Web, amplifying both their usefulness and their attack surface. Most notably, indirect prompt injection and retrieval poisoning attack the web-native carriers that survive ingestion pipelines and are very concerning. We provide OpenRAG-Soc, a compact, reproducible benchmark-and-harness for web-facing RAG evaluation under these threats, in a discrete data package. The suite combines a social corpus […]

Ver mais

arXiv:2602.16340v2 Announce Type: replace-cross Abstract: We study the implicit bias of momentum-based optimizers on homogeneous models. We first extend existing results on the implicit bias of steepest descent in homogeneous models to normalized steepest descent with an optional learning rate schedule. We then show that for smooth homogeneous models, momentum steepest descent algorithms like Muon (spectral norm), MomentumGD ($ell_2$ norm), and Signum ($ell_infty$ norm) are approximate steepest descent trajectories under a decaying learning rate schedule, proving that these […]

Ver mais

arXiv:2601.16283v1 Announce Type: new Abstract: Modern buildings are increasingly interconnected with occupancy, heating, ventilation, and air-conditioning (HVAC) systems, distributed energy resources (DERs), and power grids. Modeling, control, and optimization of such multi-domain systems play a critical role in achieving building-sector decarbonization. However, most existing tools lack scalability and physical consistency for addressing these complex, multi-scale ecosystem problems. To bridge this gap, this study presents BESTOpt, a modular, physics-informed machine learning (PIML) framework that unifies building applications, including benchmarking, […]

Ver mais

Learning-to-Defer routes each input to the expert that minimizes expected cost, but it assumes that the information available to every expert is fixed at decision time. Many modern systems violate this assumption: after selecting an expert, one may also choose what additional information that expert should receive, such as retrieved documents, tool outputs, or escalation context. We study this problem and call it Learning-to-Defer with advice. We show that a broad family of natural separated surrogates, which learn […]

Ver mais