Conventional video outpainting methods primarily focus on maintaining coherent textures and visual consistency across frames. However, they often fail at handling dynamic scenes due to the complex motion of objects or camera movement, leading to temporal incoherence and visible flickering artifacts across frames. This is primarily because they lack instance-aware modeling to accurately separate and track individual object motions throughout the video. In this paper, we propose a novel video outpainting framework that explicitly takes shadow-object pairs into consideration to enhance the temporal and spatial consistency of instances, even when they are temporarily invisible. Specifically, we first track the shadow-object pairs across frames and predict the instances in the scene to unveil the spatial regions of invisible instances. Then, these prediction results are fed to guide the instance-aware optical flow completion to unveil the temporal motion of invisible instances. Next, these spatiotemporal guidances of instances are used to guide the video outpainting process. Finally, a video-aware discriminator is implemented to enhance alignment among dynamic shadows and the extended semantics in the scene. Comprehensive experiments underscore the superiority of our approach, outperforming existing state-of-the-art methods in widely recognized benchmarks.

The need for effective unlearning mechanisms in large language models (LLMs) is increasingly urgent, driven by the necessity to adhere to data regulations and foster ethical generative AI practices. LLM unlearning is designed to reduce the impact of undesirable data influences and associated model capabilities without diminishing the utility of the model if unrelated to the information being forgotten. Despite growing interest, much of the existing research has focused on varied unlearning method designs to boost effectiveness and efficiency. However, the inherent relationship between model weights and LLM unlearning has not been extensively examined. In this paper, we systematically explore how model weights interact with unlearning processes in LLMs and we design the weight attribution-guided LLM unlearning method, WAGLE, which unveils the interconnections between'influence' of weights and'influence' of data to forget and retain in LLM generation. By strategically guiding the LLM unlearning across different types of unlearning methods and tasks, WAGLE can erase the undesired content, while maintaining the performance of the original tasks. We refer to the weight attribution-guided LLM unlearning method as WAGLE, which unveils the interconnections between'influence' of weights and'influence' of data to forget and retain in LLM generation.

Apple could unveil Gemini-powered Siri in Feb. What to expect when you're expecting the Samsung Galaxy S26 in February. Samsung's 2025 was filled with new foldables, an ultra-thin new form factor and the launch of Google's XR platform . After making some announcements at CES 2026, the company is expected to host its first Galaxy Unpacked of the year in February to introduce the Galaxy S26 lineup. Engadget will be covering Galaxy Unpacked live, and we'll most likely have hands-on coverage of Samsung's new smartphones soon after they're announced.

We already know Apple is secretly preparing its first ever foldable iPhone. But now, keen tech fans have just got a new hint at what it may look like. According to a leak from a veteran Apple analyst, Mark Gurman of Bloomberg, the'iPhone Fold' will see some major design changes. And Mr Gurman claims that anyone who tries the new foldable device will'never want to go back'. Although it has been rumoured for years, it now looks almost certain that Apple is planning to unveil the folding iPhone in September 2026.

The need for effective unlearning mechanisms in large language models (LLMs) is increasingly urgent, driven by the necessity to adhere to data regulations and foster ethical generative AI practices. LLM unlearning is designed to reduce the impact of undesirable data influences and associated model capabilities without diminishing the utility of the model if unrelated to the information being forgotten. Despite growing interest, much of the existing research has focused on varied unlearning method designs to boost effectiveness and efficiency. However, the inherent relationship between model weights and LLM unlearning has not been extensively examined. In this paper, we systematically explore how model weights interact with unlearning processes in LLMs and we design the weight attribution-guided LLM unlearning method, WAGLE, which unveils the interconnections between'influence' of weights and'influence' of data to forget and retain in LLM generation.

Apple unleashed some impressive gadgets last year including the iPhone 16 and the Vision Pro headset. Now, the tech giant is lining up its next set of gizmos – and the first five could arrive as soon as next month. According to reports, the company is about to unveil a new'affordable' iPhone – an alternative to the iPhone 16 which starts at 799/ 799. It's also preparing two new iPads, a new Mac and a brand new AI product, if rumours are to be believed. 'We've been off to a light start in 2025 in terms of Apple products,' says Apple news and rumours website 9to5Mac.

Additional Feedback: I think it is too strong to claim that "we also theoretically unveil the certain important mechanisms that govern the behavior of JCL." The main theoretical tool in the proposed method is an application of Jensen's inequality. There is also a section (3.3) that discusses some very basic properties of the the objective. To claim any of this as a significant "theoretical contribution" is too strong in my view. To me, the most interesting aspect of Fig2 is part (b).

Apple's Worldwide Developers Conference (WWDC) keynote is right around the corner. The festivities kick off on Monday, June 10 at 1PM ET. The keynote address is available to the public and you can watch it via Apple's event website or on the company's YouTube channel. Alternatively, you can just stay here and watch our embed. This is WWDC, so it'll be a software-focused event. Apple will most likely reveal many of the main features of iOS 18 and iPadOS 18.

Tesla is introducing a robotaxi on August 8, Elon Musk has announced on X a few hours after Reuters published a report that the automaker is scrapping its plans to produce a low-cost EV. Reuters also said that Musk's directive was to "go all in" on robotaxis built on the company's small-vehicle platform. Tesla has been promising a more affordable EV with prices expected to start at 25,000 for years, and Musk said as recently as this January that he's optimistic the model will arrive in the second half of 2025. In response to the report, the Tesla chief tweeted that "Reuters is lying (again)." He didn't clarify which part of the report was a lie, but considering he confirmed that Tesla is unveiling a robotaxi, he likely meant the news organization's claim that the company pulled the plug on a more affordable EV.

The festivities start at 12PM ET on Thursday, March 21. The best way to watch everything unfold is via the official Microsoft stream; it doesn't look like they'll have a stream on YouTube this time. This is the first Microsoft live event without former chief product officer and longtime keynote presenter Panos Panay, who departed for Amazon last year. The event is being advertised as a "new era of work", so there's going to be a major focus on the commercial space. There are plenty of rumors that Microsoft will take the opportunity to unveil a spate of new Surface devices.