The loss of soft-touch plastics may seem like a downgrade at first, but I vastly prefer uncoated plastics for long-term use: My MX Master 2S developed unsightly smooth spots on both mouse buttons where the soft-touch coating wore down, and other long-term users have reported the coatings becoming tacky over time.

The two primary mouse buttons and the scroll wheel are nearly dead-silent. Despite this, they still have clear feedback with a distinct bump and no mushy feeling. In the ratcheting mode, the mouse wheel has a noticeable bump between each scroll. In the smooth mode, the wheel has just enough resistance to be controlled easily, while still spinning freely.

The horizontal scroll wheel on the side permanently scrolls smoothly and has significantly more resistance than the primary scroll wheel. This allows for greater control of the wheel and helps protect against accidental scrolling when moving your thumb. Scrolling is incredibly smooth, without any catching or scratchiness, and the ridged texture of the aluminum wheel feels comfortable and easy to manipulate. Clicking the wheel is still fairly loud, like any mouse. The three side buttons and the top button have a muted click, and it’s not disruptive.

Despite weighing 150 grams, this mouse is easy to glide around a surface. The feet are smooth, sliding easily while still having enough friction for precise control, and the sculpted shape makes it easy to move and lift. While the high weight means it won’t be ideal for competitive, high-intensity gaming, the shape is preferable for longer sessions, whether you’re grinding out dungeons or slogging through spreadsheets.

The sensor is capable of up to 8,000 dots per inch, but most people will likely leave the sensitivity well below that. I didn’t have any issues with input registration, skipping, or shaking, and the sensor felt incredibly precise on all surfaces, even frosted and transparent glass.

Gesture Controls and Haptics

Photograph: Henri Robbins

The MX Master 4’s gesture controls are one of its most prominent features. These have been featured on every generation of the MX Master line, with only small changes across generations. The gesture control button was previously located on the bottom of the thumb rest, hidden underneath the rubber surface, but has now been moved to a standard button on the side, which I found more comfortable and natural to use.

Samsung makes everything from smartphones and gaming monitors, to smart TVs and dishwashers. I’m always looking for a sale (I’m assuming you are, too), and I’ve found the best Samsung promo codes and special offers to help you save big on your most important tech purchases. At WIRED, we often review the South Korean company’s products, especially Samsung’s vast lineup of Galaxy smartphones, and I’ve rounded up a bunch of Samsung coupons for (virtually) every type of shopper.

Get 10% Off With Samsung Promo Code and up to $2,100 Off Phones

Right now, Samsung has some of the best deals I’ve ever seen on their best-selling tech, and they’re about to get even better with limited-time trade-in credits, a special offer program, and bundle deals. Right now, you can get a Samsung promo code for 10% off TVs—all you have to do is register for their email newsletter. The offer is sent to your inbox and is valid through September 30.

Shop Samsung’s best coupons and offers to score major discounts (sometimes up to $2,100) on smartphones, laptops, tablets, TVs, and their latest releases. And when you buy products together that you already need, you can save a ton. This includes up to 54% select Galaxy Buds, watches and tablets when you order select products, like the Galaxy S25 Ultra.

If you’re in the market for a new Samsung phone, you can get a new Galaxy Z Fold7 for $10 less with a trade-in. Feeling nostalgic? The new spin on an old classic, the Galaxy Z Flip7 is $150 off or up to $700 off with trade-in.

Or maybe you want one of the Galaxy S25 Ultra models. Get $200 off a Galaxy S25 Ultra, you’ll get up to $940 off with instant trade-in credits, and a storage upgrade for a limited time.

Unlock a 30% Off Samsung Promo Code With Offer Programs, Plus a $100 Referral Code

One of the hottest Samsung promo codes is a whopping 30% discount for government employees, first responders, military personnel, and educators. Samsung also has offer programs, meaning you can combine your promo code discount with most other offers to increase discounts. Get a pal involved for more savings—when a friend uses your referral code to make a purchase at Samsung.com, they’ll get 5% off their purchase (up to $250 off) and you’ll get up to $100 off per order (with the potential to save $1,000 per calendar year). My insider tip is to sign up for a Samsung Rewards account and download the mobile app for even more perks, including exclusive Samsung coupons, flash sales, and updates on the newest products, like the QLED 8K, select refrigerators, and other home appliances.

Save up to 35% on These Trending Samsung TV Deals

Along with other great tech, Samsung has some seriously nice TVs. The Samsung Frame TV has been trending this year for its stylish ability to blend into your home’s decor. Plus it just feels more elevated than a regular ol’ TV and mount. Some other trending TVs this Summer have been the Q60D, S90C, and the S95D models–not only do they have instant discounts of over up to 35% ($2,100 off). Plus, there are tons of TV and home theater deals at Samsung, including a bundle offer for $7500 off when you buy a Neo QLED 4K TV with a Dolby ATMOS soundbar. If you’re in the market for a new TV, it’s worth checking out Neo QLED AI Smart TVs to score 1 year of ESPN for free (worth $299).

You can also take advantage of their Trade-In Recycling Program for up to $200 off when you trade in your old TV—any brand, any size. When your new one is delivered, Samsung will handle recycling the old one, so you can enjoy your upgrade.

$169 Off With Samsung Promo Code or a $400 Gift Card on Appliances

Although here at WIRED we mostly cover Samsung’s traditional AV tech, they also make top-of-the-line kitchen and home appliances. During the Buy More Save More Event (through December 4), you can get up to 40% off high-tech Samsung kitchen appliances along with free 3-day rush shipping. When you buy any two qualifying Samsung Appliances, you’ll receive a $100 Samsung Prepaid Mastercard; if you buy three qualifying Samsung Appliances, you’ll receive a $300 Samsung Prepaid Mastercard; and when you purchase four or more appliances, you’ll get a Samsung Prepaid Mastercard for $400.

In addition to Buy More Save More discounts, other eye-catching deals include an extra $600 off a top freezer refrigerator when you buy a Bespoke 4-Door Flex Refrigerator with AI Family Hub + AI Vision, and $1,100 off the Bespoke 4-Door French Door Refrigerator with Beverage Center. Special offers also include free installation service, plus Samsung will haul away your old appliances and recycle them, while you get a $50 energy rebate. This futuristic fridge is basically also an iPad, with an AI Family Hub with the large screen and changeable door panels. Plus, there’s AI Vision inside, so you always know what’s inside (and what you need to buy at the store). And the Beverage Center has an internal dispenser or a built-in AutoFill Water Pitcher to get cold, crisp water whenever you want it, whichever way you want.

You can save an extra $170 with code SAVE169 at checkout, plus you’ll get $50 in Samsung Rewards (equal to 10,000 bonus points) with your purchase. And right now, you can get $1,000 off a Bespoke Smart Slide-In Electric Range. This range is straight from a The Jetsons fantasy, with an AI Home LCD display, which is pretty much a kitchen robot helper that gives you personalized recipe recommendations, the ability to search for and follow video recipes, and access your favorite apps so you can see who’s at your door through your video doorbell, and more. There’s also a Smart Oven Camera inside, meaning you can check on meals as they cook from anywhere and even share time-lapse videos to show off your skills.

Stay up to Date on all Things Samsung at WIRED

WIRED also has guides to help determine which Galaxy S24 phone is best for you and how to set up your Samsung Galaxy S25 to ensure you’re getting the most out of its features, as well as advice on which Galaxy S24 series accessories, like cases, chargers, and power banks, are worth the money.

Us nerds here at WIRED also follow CES (sort of the Coachella for tech nerds) for all the updates on tech (almost) no one asked for, and Samsung’s bi-annual Galaxy Unpacked event, where they show off its newest toys. We have a lot of opinions about Samsung’s foldable Galaxy Z Flip6 and Z Fold6 phones. We are also patiently awaiting new releases of Galaxy Tab tablets, a new line of Galaxy Buds Pro 3 wireless earbuds, and a new series of the Galaxy Watch, with a new design and improved sensors for health

A new framework for generative diffusion models was developed by researchers at Science Tokyo, significantly improving generative AI models. The method reinterpreted Schrödinger bridge models as variational autoencoders with infinitely many latent variables, reducing computational costs and preventing overfitting. By appropriately interrupting the training of the encoder, this approach enabled development of more efficient generative AI, with broad applicability beyond standard diffusion models.

Diffusion models are among the most widely used approaches in generative AI for creating images and audio. These models generate new data by gradually adding noise (noising) to real samples and then learning how to reverse that process (denoising) back into realistic data. A widely used version, the score-based model, achieves this by the diffusion process connecting the prior to the data with a sufficiently long-time interval. This method, however, has a limitation that when the data differs strongly from the prior, the time intervals of the noising and denoising processes become longer, which causes slowing down sample generation.

Now, a research team from Institute of Science Tokyo (Science Tokyo), Japan, has proposed a new framework for diffusion models that is faster and computationally less demanding. They achieved this by reinterpreting Schrödinger bridge (SB) models, a type of diffusion model, as variational autoencoders (VAEs).

The study was led by graduate student Mr. Kentaro Kaba and Professor Masayuki Ohzeki from the Department of Physics at Science Tokyo, in collaboration with Mr. Reo Shimizu (then a graduate student) and Associate Professor Yuki Sugiyama from the Graduate School of Information Sciences at Tohoku University, Japan. Their findings were published in the Physical Review Research on September 3, 2025.

SB models offer greater flexibility than standard score-based models because they can connect any two probability distributions over a finite time using a stochastic differential equation (SDE). This supports more complex noising processes and higher-quality sample generation. The trade-off, however, is that SB models are mathematically complex and expensive to train.

The proposed method addresses this by reformulating SB models as VAEs with multiple latent variables. “The key insight lies in extending the number of latent variables from one to infinity, leveraging the data-processing inequality. This perspective enables us to interpret SB-type models within the framework of VAEs,” says Kaba.

In this setup, the encoder represents the forward process that maps real data onto a noisy latent space, while the decoder reverses the process to reconstruct realistic samples, and both processes are modeled as SDEs learned by neural networks.

The model employs a training objective with two components. The first is the prior loss, which ensures that the encoder correctly maps the data distribution to the prior distribution. The second is drift matching, which trains the decoder to mimic the dynamics of the reverse encoder process. Moreover, once the prior loss stabilizes, encoder training can be stopped early. This allows us to complete learning faster, reducing the risk of overfitting and preserving high accuracy in SB models.

“The objective function is composed of the prior loss and drift matching parts, which characterizes the training of neural networks in the encoder and the decoder, respectively. Together, they reduce the computational cost of training SB-type models. It was demonstrated that interrupting the training of the encoder mitigated the challenge of overfitting,” explains Ohzeki.

This approach is flexible and can be applied to other probabilistic rule sets, even non-Markov processes, making it a broadly applicable training scheme.