watchOS 11.3 has received its third developer beta, following the arrival of similar beta updates for the HomePod and Apple TV only hours earlier.Apple has released the third developer beta of watchOS 11.3.On Tuesday, Apple first released the third developer betas of tvOS 18.3 and HomePod OS 18.3, respectively. Now, the company has deployed a beta software release for its Apple Watch platform.The third developer beta of watchOS 11.3 increases the build number to 2S5550a, up from the previous 22S5540e, which was released on January 7, 2025. Continue Reading on AppleInsider | Discuss on our Forums

AIANYs picks for the 2025 AIA New York Design Awards winners were just announced from the Center for Architecture in Manhattan. A total of 24 projects representing the very best contributions made by members to the urban environment locally and abroad were selected by a jury whose 2025 members included ICON's Melodie Yashar, Yale's Anna Dyson, and Lucas ter Hall of Studio RAP.Projects were divided into four categories (Architecture, Interiors, Projects, and Urban Design), with Honor, Merit, and Citation designations given across each.Each of the winners will next be honored as part of the chapters April Honors and Awards Luncheon. The Center for Architecture will also be hosting an exhibition of the winning projects.ARCHITECTURE CATEGORY

Weve been hearing about theUntil Dawnmovie for a while now, and a few months ahead of its April 25 release, weve finally also got our first look at it. Sony has released a first look teaser that reveals new details on the movie and showcases several glimpses of what it will have in store. Check it out below.Interestingly, theUntil Dawnmovie isnt adapting the game itself. Though set in the same universe, it will tell a new story focused on a new cast. It will also take on a rather unique structure, with characters returning to life every time they die to be put through a new threat and, according to director David F. Sandberg, a new genre of horror.Sadly for fans of the game, it doesnt look like there will be a lot of overlap where the films content is concerned, but if nothing else, as previously confirmed, Peter Stormare will reprise his role of Dr. Hill from the game.Last year, leaks claimed that a sequel toUntil Dawnwas in development at Firesprite. Read more on that through here.

An artists concept of a white dwarf orbiting a supermassive black hole and shedding matter NASA / Aurore Simonnet, Sonoma State UniversityA supermassive black hole in a distant galaxy is once again surprising scientists.In 2018, a black hole called 1ES 1927+654, located about 270 million light-years away from Earth, showed its first signs of mysterious behavior. Its coronathe billion-degree cloud of plasma that envelopes itsuddenlydisappeared and reappeared. Now, a team of scientists at MIT have observed another novelty: X-ray radiation is emanating in pulses from the black hole with increasing frequency.Between 2022 and 2024, these flashes shifted from happening roughly every 18 minutes to occurring every 7.1 minutes. Just as its 2018 disappearing act was unprecedented at the time, the black holes current behavior marksIt was exciting in and of itself just to find these oscillations, because its only one of a handful [of supermassive black holes that do this], says lead researcher Megan Masterson, an astrophysicist at MIT, to New Scientists Alex Wilkins. But I think the most exciting thing to us was that the oscillation periodhow fast these oscillations were happeningwas changing on human-observable timescales, which is not usually what we see around supermassive black holes.The researchers presented their work on Monday at a meeting of the American Astronomical Society. The findings are set to be published in Nature in early February, and a pre-print paper, which has not yet been peer-reviewed, is currently available. Radio images show the black hole seemingly emitting pulses of plasma from both sides. NSF / AUI / NSF NRAO / Meyer at al. 2025The scientists say the most likely culprit behind the mysterious pulsing is a white dwarfa dense stellar remnant about the size of the Earththats teetering precariously close to the black hole. The white dwarf could be shedding matter and triggering the pulses observed by the researchers. As it circles closer and closer to the black hole, they hypothesize, those pulses speed up.Now, the white dwarf might be orbiting within millions of miles of the black holes event horizonthe point beyond which no matter or light can escape.This would be the closest thing that we know of around any black hole, Masterson says in a statementTo skirt so closely around the black hole without being swallowed up is a feat that other cosmic objects, such as a small black hole or a regular star, would not be able to achieve, per a statement from NASA. This helped give the team confidence that they were dealing with a white dwarf.As the black hole pulls the stellar remnant inward, the white dwarf loses matter from its outermost layer. The shedding of that material, the researchers say, provides a kick-back force in the opposite direction, preventing the white dwarf from falling into the black hole.While there could be other explanations for the mysterious X-ray patterns, such as oscillations in the black holes corona, those are less understood. But to confirm their white dwarf hypothesis, the MIT team will need to detect gravitational waves emanating from the black hole. To do so, theyll have to wait for the 2035 launch of LISA (Laser Interferometer Space Antenna), the first space-based observatory dedicated to detecting the gravitational waves produced by black holes and other powerful phenomena in the universe.But if astronomers determine a white dwarf is indeed causing the pulses, it could make the site more desirable for LISA to observe. Everyone really wants to see a white dwarf get eaten by a black hole with LISA, Matt Nicholl, an astrophysicist at Queens University Belfast in Northern Ireland who was not involved with the work, tells New Scientist. It helps tremendously if we have a few candidates.In the meantime, astronomers will keep their telescopes pointed on the black hole. The one thing Ive learned with this source is to never stop looking at it, because it will probably teach us something new, Masterson says in the statement. The next step is just to keep our eyes open.Get the latest stories in your inbox every weekday.Filed Under: Astronomers, Astronomy, Black Holes, New Research, Outer Space, telescope, universe

Hallucinations are a persistent problem in healthcare. Here's how GSK is using test-time compute scaling to improve its gen AI systems.Read More

Elon Musks acquisition of Twitter has resulted in a federal lawsuit by the Securities and Exchange Commission alleging that he broke securities laws with a late disclosure, and saved $150 million in the process. Before Musk agreed to buy Twitter for $44 billion, before he tried to back out of that deal, before he was forced to go through with it, and before he changed its name to X, he started by acquiring a substantial stake in the company but didnt reveal that fact until weeks later.The only problem, as the SEC pointed out then, is that by the time he disclosed that stake, it was outside the agencys required 10-day window. They claim that he shouldve filed his paperwork by March 24th, 2022, instead of when he actually did, on April 4th (and then again on April 5th). During that period, they say he purchased more than $500 million in shares of the company.However, with only a few days left before the Trump administration takes over and installs a new head of the SEC (along with Elon Musk reportedly snagging an office in the White House complex), its unclear how far the lawsuit will go.The SEC claims Musk cost investors at least $150 million due to the late disclosure and that he harmed any investors who sold stock between March 25th, 2022, and April 1st, 2022. Its lawsuit is seeking the money Elon made as a result of holding off on the disclosure, as well as a civil penalty and other punishments.

Mathematical reasoning has long been a significant challenge for Large Language Models (LLMs). Errors in intermediate reasoning steps can undermine both the accuracy and reliability of final outputs, which is particularly problematic for applications requiring precision, such as education and scientific computation. Traditional evaluation methods, like the Best-of-N (BoN) strategy, often fail to capture the intricacies of reasoning processes. This has led to the development of Process Reward Models (PRMs), which aim to provide detailed supervision by evaluating the correctness of intermediate steps. However, building effective PRMs remains a difficult task, primarily due to challenges in data annotation and evaluation methodologies. These obstacles highlight the need for models that better align with robust, process-driven reasoning.The Alibaba Qwen Team recently published a paper titled Lessons of Developing Process Reward Models in Mathematical Reasoning. Alongside this research, they introduced two PRMs with 7B and 72B parameters, part of their Qwen2.5-Math-PRM series. These models address significant limitations in existing PRM frameworks, employing innovative techniques to improve the accuracy and generalization of reasoning models.Central to their approach is a hybrid methodology that combines Monte Carlo (MC) estimation with a novel LLM-as-a-judge mechanism. This integration enhances the quality of step-wise annotations, making the resulting PRMs more effective in identifying and mitigating errors in mathematical reasoning. The models have demonstrated strong performance on benchmarks like PROCESSBENCH, which tests a models ability to pinpoint intermediate reasoning errors.Technical Innovations and BenefitsThe Qwen teams methodology involves generating multiple solutions for mathematical problems using fine-tuned LLMs and evaluating the correctness of each step through a dual approach. This method addresses the limitations of traditional MC estimation, which often produces inaccurate labels due to its reliance on future outcomes.Key innovations include:Consensus Filtering: This mechanism retains data only when both MC estimation and LLM-as-a-judge agree on step correctness, significantly reducing noise in the training process.Hard Labeling: Deterministic labels, verified by both mechanisms, enhance the models ability to distinguish valid from invalid reasoning steps.Efficient Data Utilization: By combining MC estimation with LLM-as-a-judge, the consensus filtering strategy ensures high-quality data while maintaining scalability. This approach enables the development of PRMs that perform well even with smaller datasets.These innovations facilitate the creation of PRMs that are not only accurate but also robust, making them suitable for applications such as automated tutoring and complex problem-solving.Results and InsightsThe Qwen2.5-Math-PRM models demonstrated strong results on PROCESSBENCH and other evaluation metrics. For example, the Qwen2.5-Math-PRM-72B model achieved an F1 score of 78.3%, surpassing many open-source alternatives. In tasks requiring step-wise error identification, it outperformed proprietary models like GPT-4-0806.The consensus filtering approach played a crucial role in improving training quality, reducing data noise by approximately 60%. While MC estimation alone can be helpful, it is insufficient for accurately labeling reasoning steps. Combining MC estimation with LLM-as-a-judge significantly enhanced the models ability to detect errors, as reflected in improved PROCESSBENCH scores.The Qwen2.5-Math-PRM series also emphasized step-level evaluation over outcome-based BoN strategies. This shift addressed the shortcomings of earlier models, which often prioritized final answers at the expense of reasoning accuracy.ConclusionThe introduction of the Qwen2.5-Math-PRM models represents meaningful progress in mathematical reasoning for LLMs. By addressing challenges in PRM development, such as noisy data annotation and process-to-outcome biases, the Alibaba Qwen Team has provided a practical framework for improving reasoning accuracy and reliability. These models not only outperform existing alternatives but also offer valuable methodologies for future research. As PRMs continue to advance, their application in broader AI contexts promises to enhance the reliability and effectiveness of machine reasoning systems.Check out the Paper and Models on Hugging Face. All credit for this research goes to the researchers of this project. Also,dont forget to follow us onTwitter and join ourTelegram Channel andLinkedIn Group. Dont Forget to join our65k+ ML SubReddit.(Promoted) Asif RazzaqAsif Razzaq is the CEO of Marktechpost Media Inc.. As a visionary entrepreneur and engineer, Asif is committed to harnessing the potential of Artificial Intelligence for social good. His most recent endeavor is the launch of an Artificial Intelligence Media Platform, Marktechpost, which stands out for its in-depth coverage of machine learning and deep learning news that is both technically sound and easily understandable by a wide audience. The platform boasts of over 2 million monthly views, illustrating its popularity among audiences. Meet 'Height':The only autonomous project management tool (Sponsored)

Fine-Tuning a Pre-trained LLM for Sentiment Classification 0 like January 14, 2025Share this postAuthor(s): Dimitris Effrosynidis Originally published on Towards AI. Optimizing results with minimal effortThis member-only story is on us. Upgrade to access all of Medium.Image by author.In a previous tutorial, Traditional vs. Generative AI for Sentiment Classification, we predicted the sentiment of product reviews from the Flipkart Customer Review dataset.We compared several methods:Logistic Regression with TF-IDF: A simple yet effective baseline using term-frequency-based features for classification.Logistic Regression with Pretrained Embeddings: Utilize advanced embedding models like all-MiniLM-L6-v2 to generate semantic representations for training a classifier.Zero-shot Classification: Perform classification without labeled data by leveraging cosine similarity between document and label embeddings.Generative Models: Explore generative language models like Flan-T5, which classify text by generating responses based on a prompt.Task-Specific Sentiment Models: Leverage fine-tuned sentiment models like juliensimon/reviews-sentiment-analysis for domain-specific performance.Here are the results of that experiment:Image by author.In this tutorial, we will fine-tune the Task-Specific Sentiment Model (juliensimon/reviews-sentiment-analysis) and see if its 0.79 accuracy will improve.To get the complete code, visit my GitHub portfolio.If you run this code on Google Colab (or any other cloud platform), please ensure that all necessary dependencies are installed.Run the following code block to install the required packages:%%capture!pip install datasets transformers sentence-transformers evaluateWe will use the same dataset and the same pre-processing as the previous tutorial.import pandas as pdimport numpy as Read the full blog for free on Medium.Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming asponsor. Published via Towards AITowards AI - Medium Share this post

Right now, Amazon and Best Buy are offering the current generation Apple iPad Mini (A17 Pro) for only $399.99 shipped after a $100 (20% off) discount. That matches the best deal I saw during Black Friday 2024. If you're looking for an iPad that's both powerful and pocketable, this is your best pick.2024 Apple iPad Mini (A17 Pro)iPad Mini (A17 Pro)The iPad Mini may be the most diminutive tablet of the bunch, but that doesn't mean it's hindered in any way. This is the most compact and portable iPad measuring 7.7" long, 5.3" wide, and only 0.25" thin. The compact form factor still packs in a gorgeous 8.3" Liquid Retina display with True Tone, P3 Wide Color, and ultra-low reflectivity. The new iPad Mini was released in October of 2024 and features some pretty significant improvements over its predecessor, including a more powerful A17 Pro processor, double the base storage capacity, an updated faster Wi-Fi 6e protocol, a USB Type-C port that offers faster charging, and compatibility with the newer Apple Pencil Pro and Apple Pencil (USB-C), and upcoming support for Apple Intelligence.Looking for More iPad Resources?If you're not sure which iPad is best for you, we have an iPad guide which details which iPad is ideal for which use case. If you intend want to get an iPad for schoolwork, we have an iPad guide for students as well. If you're looking for options outside of iOS, check out the best Android tablets of 2025.Why Should You Trust IGN's Deals Team?IGN's deals team has a combined 30+ years of experience finding the best discounts in gaming, tech, and just about every other category. We don't try to trick our readers into buying things they don't need at prices that aren't worth buying something at. Our ultimate goal is to surface the best possible deals from brands we trust and our editorial team has personal experience with. You can check out our deals standards here for more information on our process, or keep up with the latest deals we find on IGN's Deals account on Twitter.Eric Song is the IGN commerce manager in charge of finding the best gaming and tech deals every day. When Eric isn't hunting for deals for other people at work, he's hunting for deals for himself during his free time.

Apple Watch Ultra 3 is coming later this year. It will reportedly feature satellite connectivity, 5G cellular, and high blood pressure detection. But the thing that might win me over isnt hardware, and it isnt even actually new at all.Ultra-exclusive watch faces are a big drawLast fall, Apple Watch Series 10 arrived with a display thats just as large as what the Ultra offersbut for nearly half the price.I bought the Series 10 and have worn it gladly every day since. But theres one big shortcoming thats been a disappointment: the lack of watch faces that take advantage of its large display.Apple Watch Ultra offers two such faces:Modular UltraWayfinderThe former face is especially of interest to me.Modular Ultra lets you set seven complications, plus six options for customizing how the time is displayed, and even a special area around the edges to show more data.Its the face I want on my Series 10, but cant get.And its the face that might just get me to switch to the Apple Watch Ultra 3 this year.Before I get there though, let me share my thinking that led to the Series 10 purchase in the first place.Apple Watch Series 10 brought the hypeWhen Apple unveiled its new Watch lineup in September, the Apple Watch Ultra 3 was notably absent. Instead of releasing a new Ultra model, Apple just gave the existing Ultra 2 a black color option.As beautiful as the black Ultra 2 is, I was in the market for a new Watch, and buying a year-old model wasnt in the cards.If the Apple Watch Series 10 had been a minor update, I could have been tempted by the Ultra 2. But instead, Series 10 brought:a thinner designlarger screensS10 chipfaster chargingmore frequent always-on refreshesand wide-angle OLEDWhat did the Apple Watch Ultra get?A new color.Its easy to see why the Series 10 was my quick choice.Switching to Apple Watch Ultra 3 for one (old) featureAs much as Ive enjoyed my Apple Watch Series 10, the lack of large screen-optimized watch faces continues to bug me.So much so that Im considering switching to the Ultra 3 this year.The new Ultra 3 features sound solid. Im sure more details will trickle out over the months leading up to September. We should get a new chip, maybe better battery life even.But the main draw of the Ultra 3 for me will be that Modular Ultra face.Watch faces are the heart of watchOS. Your watch face is more important than, say, your iPhones Home Screen.In most cases, the vast majority of Watch interactions involve the watch face.Glancing at the time, at widgets, launching appsall of these happen from the watch face.So getting the best possible face, one that takes advantage of a large display, is especially importantat least for me.Id love to hear that Apple is going to bring new faces to watchOS 12 that power up the Series 10 model. If that happens, I may stick with my current Watch.But if nothing changes, the Apple Watch Ultra 3 and its exclusive watch faces will be calling my name come September.What factors influence your Apple Watch purchase decisions? Let us know in the comments.Best Apple Watch accessoriesAdd 9to5Mac to your Google News feed. FTC: We use income earning auto affiliate links. More.Youre reading 9to5Mac experts who break news about Apple and its surrounding ecosystem, day after day. Be sure to check out our homepage for all the latest news, and follow 9to5Mac on Twitter, Facebook, and LinkedIn to stay in the loop. Dont know where to start? Check out our exclusive stories, reviews, how-tos, and subscribe to our YouTube channel