The rise of large language models (LLMs) has significantly influenced the quality of information in decision-making systems, leading to the prevalence of AI-generated content and challenges in detecting misinformation and managing conflicting information, or "inter-evidence conflicts." This study introduces a method for generating diverse, validated evidence conflicts to simulate real-world misinformation scenarios. We evaluate conflict detection methods, including Natural Language Inference (NLI) models, factual consistency (FC) models, and LLMs, on these conflicts (RQ1) and analyze LLMs' conflict resolution behaviors (RQ2). Our key findings include: (1) NLI and LLM models exhibit high precision in detecting answer conflicts, though weaker models suffer from low recall; (2) FC models struggle with lexically similar answer conflicts, while NLI and LLM models handle these better; and (3) stronger models like GPT-4 show robust performance, especially with nuanced conflicts. For conflict resolution, LLMs often favor one piece of conflicting evidence without justification and rely on internal knowledge if they have prior beliefs.

Soon you could see Waymo self-driving tech in Hyundai cars. The autonomous driving tech developer Waymo said this week that it would partner with the Korean automaker Hyundai to equip a fleet of its electric vehicles with self-driving technology. The vehicles, modified Ioniq 5s, will hit the road as part of Waymo's self-driving ride-hail service in late 2025, the companies said. In a statement, Hyundai Motor Company president and global COO José Muñoz called the agreement a "first step" in the two firms' partnership. "We are actively exploring additional opportunities for collaboration," he said--opening up the possibility that Waymo self-driving tech could one day be installed on Hyundai passenger vehicles.

The United States military has struck a number of cities in Yemen, including the capital, Sanaa, and the key port city of Hodeidah. Forces from the US Central Command (CENTCOM), the military command responsible for US forces in the Middle East, "conducted strikes on 15 Houthi targets in Iranian-backed Houthi-controlled areas of Yemen today", it said on X on Friday. Four strikes targeted Sanaa and seven hit Hodeidah, according to the Houthi-run Al Masirah TV network. Correspondents with the AFP news agency also reported hearing loud explosions in both cities. The Hodeidah strikes hit the airport and the Katheib area, which has a Houthi-controlled military base, Al Masirah said.

In a future conflict, American troops will direct the newest war machines not with sprawling control panels or sci-fi-inspired touchscreens, but controls familiar to anyone who grew up with an Xbox or PlayStation in their home. Over the past several years, the US Defense Department has been gradually integrating what appear to be variants of the Freedom of Movement Control Unit (FMCU) handsets as the primary control units for a variety of advanced weapons systems, according to publicly available imagery published to the department's Defense Visual Information Distribution System media hub. Those systems include the new Navy Marine Corps Expeditionary Ship Interdiction System (NMESIS) launcher, a Joint Light Tactical Vehicle–based anti-ship missile system designed to fire the new Naval Strike Missile that's essential to the Marine Corps' plans for a notional future war with China in the Indo-Pacific; the Army's new Maneuver-Short Range Air Defense (M-SHORAD) system that, bristling with FIM-92 Stinger and AGM-114 Hellfire missiles and a 30-mm chain gun mounted on a Stryker infantry fighting vehicle, is seen as a critical anti-air capability in a potential clash with Russia in Eastern Europe; the Air Force's MRAP-based Recovery of Air Bases Denied by Ordnance (RADBO) truck that uses a laser to clear away improvised explosive devices and other unexploded munitions; and the Humvee-mounted High Energy Laser-Expeditionary (HELEX) laser weapon system currently undergoing testing by the Marine Corps. The FMCU has also been employed on a variety of experimental unmanned vehicles, and according to a 2023 Navy contract, the system will be integral to the operation of the AN/SAY-3A Electro-Optic Sensor System (or "I-Stalker") that's designed to help the service's future Constellation-class guided-missile frigates track and engage incoming threats. Produced since 2008 by Measurement Systems Inc. (MSI), a subsidiary of British defense contractor Ultra that specializes in human-machine interfaces, the FMCU offers a similar form factor to the standard Xbox or PlayStation controller but with a ruggedized design intended to safeguard its sensitive electronics against whatever hostile environs American service members may find themselves in.

Although gradient descent (GD) almost always escapes saddle points asymptotically [Lee et al., 2016], this paper shows that even with fairly natural random initialization schemes and non-pathological functions, GD can be significantly slowed down by saddle points, taking exponential time to escape. On the other hand, gradient descent with perturbations [Ge et al., 2015, Jin et al., 2017] is not slowed down by saddle points--it can find an approximate local minimizer in polynomial time. This result implies that GD is inherently slower than perturbed GD, and justifies the importance of adding perturbations for efficient non-convex optimization. While our focus is theoretical, we also present experiments that illustrate our theoretical findings.

Data augmentation is a ubiquitous technique for increasing the size of labeled training sets by leveraging task-specific data transformations that preserve class labels. While it is often easy for domain experts to specify individual transformations, constructing and tuning the more sophisticated compositions typically needed to achieve state-of-the-art results is a time-consuming manual task in practice. We propose a method for automating this process by learning a generative sequence model over user-specified transformation functions using a generative adversarial approach. Our method can make use of arbitrary, non-deterministic transformation functions, is robust to misspecified user input, and is trained on unlabeled data. The learned transformation model can then be used to perform data augmentation for any end discriminative model. In our experiments, we show the efficacy of our approach on both image and text datasets, achieving improvements of 4.0 accuracy points on CIFAR-10, 1.4 F1 points on the ACE relation extraction task, and 3.4 accuracy points when using domain-specific transformation operations on a medical imaging dataset as compared to standard heuristic augmentation approaches.

We consider a large-scale matrix multiplication problem where the computation is carried out using a distributed system with a master node and multiple worker nodes, where each worker can store parts of the input matrices. We propose a computation strategy that leverages ideas from coding theory to design intermediate computations at the worker nodes, in order to optimally deal with straggling workers. The proposed strategy, named as polynomial codes, achieves the optimum recovery threshold, defined as the minimum number of workers that the master needs to wait for in order to compute the output. This is the first code that achieves the optimal utilization of redundancy for tolerating stragglers or failures in distributed matrix multiplication. Furthermore, by leveraging the algebraic structure of polynomial codes, we can map the reconstruction problem of the final output to a polynomial interpolation problem, which can be solved efficiently. Polynomial codes provide order-wise improvement over the state of the art in terms of recovery threshold, and are also optimal in terms of several other metrics including computation latency and communication load. Moreover, we extend this code to distributed convolution and show its order-wise optimality.

Neural Information Processing Systems http://nips.cc/

Gen. Jack Keane joins'Fox Report' to discuss the escalating tensions in the Middle East amid fears of a wider war. While much of the world has eyes on Israel's battles with Hezbollah and Hamas, the U.S. Navy has its sights set on another of Iran's proxies, the Yemeni Houthi rebels. With a mission to keep international waterways at peace, the Navy now finds itself fending off attacks from the shadowy gang of pirates who have gone from arming themselves with assault rifles, pickup trucks and motorboats – to a seemingly unending supply of drones, missiles and other weaponry. The Houthis often attack unarmed Western ships carrying goods through the Red Sea and the Gulf of Aden – while the U.S. has responded in kind with drone attacks on Yemen. ISRAELI AIR FORCE STRIKES HOUTHI TARGETS IN YEMEN WITH'EXTENSIVE' OPERATION That's led to perilous waters along a trade route that typically sees some 1 trillion in goods pass through it, as well as shipments of aid to war-torn Sudan and the Yemeni people.

Obtaining enough labeled data to robustly train complex discriminative models is a major bottleneck in the machine learning pipeline. A popular solution is combining multiple sources of weak supervision using generative models. The structure of these models affects the quality of the training labels, but is difficult to learn without any ground truth labels. We instead rely on weak supervision sources having some structure by virtue of being encoded programmatically. We present Coral, a paradigm that infers generative model structure by statically analyzing the code for these heuristics, thus significantly reducing the amount of data required to learn structure. We prove that Coral's sample complexity scales quasilinearly with the number of heuristics and number of relations identified, improving over the standard sample complexity, which is exponential in n for learning n