Ecological concerns and the pressures of the modern world mean designers no longer have the luxury of seeing their work as a form of self-expression, Japanese designer Naoto Fukasawa tells Dezeen in this interview."Many designers in the past believed that design was merely a medium for self-expression, but we are now in an era where it is crucial to carefully assess whether such an approach is truly appropriate," Fukasawa said."Success means creating something that brings joy to people, while failure occurs when the design contributes to harming the natural environment or disrupting the peace of people's daily lives."Fukasawa has produced furniture and product designs for companies such as Herman Miller, Alessi, B&B Italia, Emeco, Magis, Issey Miyake and Hay.Fukasawa's wall-mounted CD player for MUJI is possibly his best known design. Photo by Hidetoyo SasakiHe is perhaps best known for his wall-mounted CD player for Muji, which is now part of the permanent collections of the Museum of Modern Art in New York and the Victoria and Albert Museum in London.The first major exhibition in the US celebrating the designer's career has now opened at The Philadelphia Museum of Art (PMA), curated by the institute's curator and historian of decorative arts, design and craft, Colin Fanning.Titled Naoto Fukasawa: Things in Themselves, it looks back at the last 25 years of Fukasawa's work.The show is too small to exhibit more than the 1,000 designs he has created, so Fukasawa said it provided him with "a great opportunity to focus on symbolic works that aligned with the key concepts behind my designs".Naoto Fukasawa: Things in Themselves is at the Philadelphia Museum of Art until 20 April. Photo by Sarah Croop"I realised that both my ideas and the quality of my designs have remained consistent," he said. "I am simply given ideas or insights from something, and then I transform those into a tangible image."At the start of a project, Fukasawa said, he asks a client or collaborator what their best-selling product is."The question is essentially asking which product design is loved by the most people," he explained. "If the answer aligns with what I feel, I will be convinced and put in the effort to create great designs with the client."From then on, his process is meticulous."I start by sketching, turning it into a drawing, creating a hand-made, highly accurate shape model," he said."Then [I] adjust and modify the design by adding technology and regulations, refining it further through the process of trial and error, creating functional prototypes, and repeatedly checking whether it matches the original vision. Ideas and execution are inseparable."Read: Naoto Fukasawa inserts Issey Miyake store into 132-year-old Kyoto townhouseIn 2025, the first question designers must ask themselves is whether or not to create more stuff in the first place, Fukasawa believes."Good design should be something that is loved for a long time," he said. "Designers and humanity in general must avoid creating things that are deemed undesirable in the context of life itself."While he still believes that "design is essential for living well", Fukasawa argues that designers are now required to think harder about their work than in the past."It's not just about creating good designs, but about making the right ones," he said. "I believe that design should serve as a tool needed by people, not merely as a source of stimulation.""I hope designers will create with humility, crafting things that blend into the background of everyday life."Fukasawa designed the Cha teapot for AlessiDesign fairs have come under increasing scrutiny in recent years from some commentators who argue they encourage unnecessary production.But Fukasawa believes they still have value, as long as they do more than simply exhibit products."Their value lies in the ability to physically engage with products and experience, whether they improve usability and enhance daily life," he said."I think fairs should be places where the value becomes clear through experience. They shouldn't just be trade shows."Nevertheless, he expressed some reservations about brands pouring large amounts of resources into fully immersive installations at various design weeks."While it's not a bad thing to have a space where brands that offer experiential products come together, I don't think it's ideal to spend a huge amount of money on building stands that are then dismantled and discarded after a week," he said.Grande Papilio chair was designed by Naoto Fukasawa for B&B Italia. Photo courtesy of B&B ItaliaFukasawa said his principles will continue to apply as artificial intelligence becomes more commonly used in the design process."AI has already become an indispensable tool in bringing images to life," he said."While the benefits of AI are clear, recent trends showing a growing interest in folk crafts and handmade items suggest that people are increasingly drawn to things created directly by the human body, not just driven by convenience, reflecting a deeper sense of attachment and emotion."Other designers recently interviewed by Dezeen include art-design duo the Haas Brothers and Dutch designer Sabine Marcelis.The photography is courtesy of the PMA unless otherwise stated.Naoto Fukasawa: Things in Themselves is on show at the Philadelphia Museum of Art until 20 April. See Dezeen Events Guidefor more architecture and design events around the world.The post Design should "blend into the background of everyday life" says Naoto Fukasawa appeared first on Dezeen.

Data journalist and illustrator Mona Chalabi has worked with architecture studio Situ to examine the proposed international crime of domicide in Patterns of Life, an installation at the Smithsonian Design Triennial at New York's Cooper Hewitt museum.Presented under the triennial's theme of "Making Home", Patterns of Life explores the impacts of the systematic destruction of housing by focusing on three real-world examples, represented by three architectural models.The models recreate in dimensionally precise detail the homes of Osman and his wife in Manbij, Syria, damaged in 2016; Basim and his family in Mosul, Iraq, destroyed in 2015; and an unnamed woman and her son in Gaza, Palestine, destroyed in 2023.The Patterns of Life installation recreates homes destroyed by domicide as architectural modelsAlongside the models, Chalabi's illustrations both situate the homes within their wider cities and provide a window into the lives that were lived within.Chalabi conceived the installation in collaboration with Situ Research, a wing of Brooklyn-based practice Situ, to put a human face on the widespread destruction in overseas conflicts that the public sees in the media.She was first approached by the triennale in 2023 and initially began developing a work around US housing and inequality before pivoting once the scale of the Israeli offensive in Gaza began to dwarf her other concerns.There is still a connection to the US, however all three of the homes featured in the exhibition were destroyed by American munitions, with a direct link proved through Situ's research.The largest of the models depicts the home of Osman and his wife in Manbij, Syria"A large part of this project was to recast how a lot of these countries are represented in US media," Situ Research deputy director Gauri Bahuguna told Dezeen. "You're not really shown these private, intimate moments of domestic life.""Another part of that is just this relentless depiction of rubble," added Chalabi. "The rubble only shows the after it doesn't really show you what's been lost.""It's so hard for us [as viewers] to look at piles and piles of grey concrete broken into stones and to visually be able to construct what was there, whereas a resident of that city will look out and they will see what was there before," she continued.The homes in Patterns of Life all have different typologies. Basim's home in Iraq is the single-family house of someone relatively affluent, while the home in Gaza is a low-rise multi-family residence and Osman's in Syria is a larger apartment building.A smaller apartment building forms the home of woman X and her son in Gaza, PalestineSitu's fabrication unit constructed the models as accurately as possible based on satellite imagery gathered by Situ Research as well as the photographs and personal testimonies of the participants that Chalabi interviewed.Inside the models, the team took a more non-literal approach. On one facade, the models are cut to reveal a cross-section of the interior, recreated with an almost dreamlike quality that aims to communicate a feeling of loss.Here, Chalabi's illustrations of domestic details remembered by her interviewees such as books, toys and heirlooms are presented as digital prints on layers of translucent silk, giving both a sense of three-dimensionality and fragility.Bahuguna said that this approach arose in an effort to combine architectural models and illustration without entering "doll house territory".Read: "Data replicates the existing systems of power" says Pulitzer Prize-winner Mona Chalabi"There were conversations about, how do you get someone to feel a bit more softness or evoke some sense of domesticity without then veering into the twee?" she said."We talked about having 3D objects, but it would have felt too twee or too much like a doll's house," added Chalabi. "There needed to be some degree of abstraction.""If it was a full replica of the home, you would almost lose sight that this had been destroyed. We wanted to communicate that something has been lost here this is an attempt to recreate memories, rather than to recreate the home itself," she continued.Fragrance also helps to evoke emotion, with peppermint oil in the Syrian home honouring Osman's memories of mint tea, lavender oil in the Iraqi home nodding to Basim's garden, and orange oil in the Palestinian home evoking surrounding orchards.Basim's home in Iraq is a single-family houseChalabi and Bahuguna hope that, after walking through many other rooms exploring housing in the USA, visitors to Patterns of Life are drawn to reflect on the country's connection to these homes overseas and their destruction.Domicide is not currently a crime under international law, but it has been on the agenda since it was recommended in a 2022 report by the UN Special Rapporteur on the Right to Adequate Housing, Balakrishnan Rajagopal.The report considers not just the deliberate destruction of homes through military conflict but through urban development or social upheaval as kinds of domicide."It felt really important to communicate that domicide isn't about the destruction of a house, it's about the destruction of a home," said Chalabi."Under international law, a home is not protected; the only protections that exist reduce home to private property," added Bahuguna. "But there is something more than just the physical property that is being destroyed."The Cooper Hewitt Smithsonian Design Museum's seventh triennial exhibition, Making Home features a wide range of works, including an installation by The Architecture of Reentry imagining the best way for individuals to reenter communities from prison and an intervention by the Black Artists + Designers Guild restocking the museum's library.The photography is by Elliot Goldstein.The Smithsonian Design Triennial is on show from November 2024 to 10 August 2025 in New York City. For more exhibitions, talks and fairs in architecture and design visit Dezeen Events Guide.Project credits:SITU Research team: Gauri Bahuguna, Martina Duque Gonzalez, Sam Rabiyah, Frederick Rapp, Bradley Samuels, Candice StrongwaterThe post Patterns of Life installation recreates homes destroyed by domicide appeared first on Dezeen.

The Nintendo Switch has lit a fire in the hearts of gamers who have discovered that they can now play their games almost anywhere, sometimes even in the most unimaginable or even inappropriate of places. Of course, that includes playing in bed, a dream come true for some gamers, at least until theyre rudely woken up to the reality of a heavy slab of plastic and metal falling on your face after youve accidentally dozed off. The Switch is a relatively lightweight device, almost as light as an iPad mini, but it will still hurt.Now imagine if that were a Steam Deck instead, a more powerful but also bulkier and heftier handheld gaming machine. Theoretically, you shouldnt be playing before bedtime in the first place, but to each their own. If you are the type to often have devices fall flat on your face and ruin your night, this accessory might just save your life from concussions, letting you comfortably hold your Steam Deck aloft, even after youve gone way past your bedtime.Designer: MechanismStrictly speaking, the actual product here isnt the harness but a rather small mount that lets you connect your Steam Deck to almost any GoPro accessory, specifically camera mounts. Its so small and simple that youd be wondering why no one has thought of this solution before. Then again, very few will probably shamefully admit to having the Steam Deck painfully drop on their face either.It just so happens that there are a ton of GoPro mounts available, given the popularity of the line of action cameras, and that includes a chest harness that is primarily intended for hands-free POV recording. With this Steam Deck Go Pro mount, though, that also becomes a way to raise your heavy handheld PC higher when youre in bed and keep it in place without having to constantly hold it. And in case you do let go, intentionally or otherwise, you can have the peace of mind that the Deck wont just fall on you.Beyond this admittedly niche use case, the mount does open up other avenues for the Steam Deck. It can be turned into an ad hoc desktop computer, for example, raising the small screen to higher levels. You can also clip the Deck on shelves, allowing you to enjoy some videos away from your desk. And yes, you can even pair this with other devices, as long as you have the right grip for it.In fact, this GoPro Mount is practically useless unless you also have the brands DeckMate grip for the Steam Deck. The mount is also technically compatible with almost any GoPro accessory, but just because you can doesnt always mean you should. For one, you still have to consider the weight of the Steam Deck, which might not be supported by other mounts. For another, there might be places and instances where using a Steam Deck might not only be inappropriate but also illegal, so best to still exercise common sense.The post Steam Deck Chest Harness will save your face from a heavy brick when in bed first appeared on Yanko Design.

Apple will launch a slate of new Macs in 2025, from laptops to desktops. Which one is best for creatives?

Following a complete outage on Saturday, the PlayStation Network is back online. Sony has apologized for the inconvenience, blaming the issues on an operational issue and confirmed that all PlayStation Plus subscribers would receive five free days of service.Unfortunately, the outrage affected several titles with online events or ongoing betas. Destiny 2s latest dungeon, Sundered Doctrine, saw its World First race hampered due to the outage. The same also applied to Capcoms latest open beta for Monster Hunter Wilds. At least theres another weekend to look forward to from February 13th to 16th.Sony hasnt revealed the nature of this issue or what caused it but stay tuned for updates. On the bright side, the company has removed mandatory PSN account linking for some of its titles on Steam. Instead, its now offering benefits for those who opt for the same, like additional outfits in The Last of Us Part 2 Remastered, Horizon Zero Dawn Remastered, and more.Network services have fully recovered from an operational issue. We apologize for the inconvenience and thank the community for their patience. All PlayStation Plus members will automatically receive an additional 5 days of service. Ask PlayStation (@AskPlayStation) February 9, 2025

After debuting with impressive numbers on Steam, Warhorse Studios Kingdom Come: Deliverance 2 has achieved a bigger concurrent player record over the weekend. Its reached 256,206 players according to SteamDB (at this time of writing) and is the 49th most-played game in Steam history based on lifetime peak totals.If that wasnt enough, the sequel sold over one million copies within 24 hours, beating its predecessor that achieved the same goal in two weeks. Kingdom Come: Deliverance 2 is a direct follow-up, continuing Henrys journey as he travels through two new maps the Bohemian Paradise and Kuttenberg.Both are impressively large with Warhorse touting double the first games content and roughly 100 hours of playtime. Theres also an unprecedented amount of freedom as players can pursue their own goals with varying consequences (and a reactive world who may or may not take kindly to them).For more details, check out our review for Kingdom Come: Deliverance 2 here. Alongside PC, its available for PS5 and Xbox Series X/S.

Monday, February 10th, 2025Posted by Jim ThackerTopaz Labs releases Gigapixel 8.2html PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN" "http://www.w3.org/TR/REC-html40/loose.dtd"Gigapixel 8s new Redefine model uses generative AI to fill in missing details in images.Originally posted on 28 October 2024. Scroll down for details of the Gigapixel 8.2 update.Topaz Labs has released Gigapixel 8, the latest major version of its AI-trained software for up-resing photos or rendered images without softening them or introducing noise.Version 8.0 adds a new generative AI-based Redefine model for sharpening and stylizing images, and updates the softwares Face Recovery model.It is also now possible to process images in the cloud, as well as on your local machine, with cloud processing charged on a credit basis.Machine-learning-trained tool enlarges images without sacrificing detailGigapixel is designed to enlarge digital images while preserving their sharpness, using machine learning techniques to fill in the missing details.Its algorithms are trained using a data set of millions of images, with Topaz Labs claiming that it can enlarge images up to 16x before image quality starts to drop visibly.Although its marketed primarily at photographers, it can also be used on rendered imagery, and has had a dedicated AI mode for CG images since version 5.3.As well as upscaling low-res renders to generate high-res images more quickly, users can remove noise from renders by upscaling them, then reducing them to their original size.Text prompts can be used to guide the output of Gigapixels new generative AI-based Redefine model, particularly when using it at high Creativity settings.Gigapixel 8.0: new generative AI-based Redefine modelTopaz Labs describes Gigapixel 8.0 as the first upscaler using generative AI models to create high-quality photorealistic results.The softwares new Redefine model uses generative AI to fill in missing detail when upscaling images to very high resolutions, or repairing damaged images.It is also surprisingly good at denoising and sharpening in some scenarios.The model comes with a Creativity setting, which determines to what extent the image is stylized: low values result in realistic sharpening, while higher values produce progressively more and more stylized results.For high Creativity settings, users can also enter text prompts to guide the content of the AI details generated, in a similar way to a conventional generative AI tool.The Redefine model currently runs on the GPU, and there are some known issues with AMD GPUs and on macOS: you can find more details in the release notes.Updates to Face RecoveryGigapixels existing Face Recovery model has been updated, with the Gen 2 model improving performance when upscaling faces seen in profile.Gen 2 also results in cleaner teeth and eyeglasses, and performs better when recovering obscure faces with little-to-no facial attributes.Option to render images in the cloud, as well as on a local machineThe other major change in Gigapixel 8 is the option to process images in the cloud as well as on your local machine.Cloud rendering is available when using the generative AI models: both Redefine and Recover, which was introduced in Gigapixel 7.1 for upscaling very low-res images.The service is priced on a credit basis, with cost depending on image size outputting larger images costs more and the AI model used.https://www.cgchannel.com/wp-content/uploads/2024/10/250210_Gigapixel82_tw.mp4Updated 10 February 2025: Topaz Labs has released Gigapixel 8.2.The update adds the option to upscale images based on their longest side, making it possible to batch process images with varying aspect ratios.It is also now possible to batch upload and batch process images in the cloud, as shown above.Topaz Labs has also improved performance of the generative AI-based Redefine model, particularly on macOS, although Mac users are still reporting issues in the release thread.The previous update, Gigapixel 8.1, added the option to order physical prints of images from within Gigapixel, via a partnership with photo printing firm WhiteWall. Price and system requirementsGigapixel is available as a standalone application for Windows 10+ and macOS 11.0+; and as a plugin for Photoshop and Lightroom Classic. The software is OpenGL-based, and supports AMD, Apple Silicon, Intel and NVIDIA GPUs.New licences of the standard edition cost $99. The new Gigapixel Pro edition for studios is subscription only and costs $499/year.Cloud render credits cost $5 for 20 credits. Credit subscriptions start at $9.99/month for 80 credits per month.Read a full list of new features in Gigapixel 8.0 to 8.2 in the online release notesHave your say on this story by following CG Channel on Facebook, Instagram and X (formerly Twitter). As well as being able to comment on stories, followers of our social media accounts can see videos we dont post on the site itself, including making-ofs for the latest VFX movies, animations, games cinematics and motion graphics projects.Latest NewsTopaz Labs releases Gigapixel 8.2Updated: AI app for upresing photos and 3D renders gets generative AI-based upscaling model; option to batch process images in the cloud.Monday, February 10th, 2025See CG benchmarks for NVIDIA's GeForce RTX 5090 and 5080 GPUsDiscover how NVIDIA's Blackwell GPUs performed in CG software like 3ds Max, Arnold, Blender, Maya, UE5 and V-Ray in our early tests. Friday, February 7th, 2025Chaos releases Beta 2 of the new V-Ray for BlenderUpdated: New free public beta build of the long-awaited Blender integration for V-Ray based on the V-Ray 7 core is now availableFriday, February 7th, 2025Review: Huion Kamvas 22 Plus pen displayThe Kamvas 22 Plus is an affordable pen display for beginners and pro artists on a budget. See how it handled 3D as well as 2D workflows.Thursday, February 6th, 2025CETA Software launches Artist AccessCheck out the new time-tracking and forecasting tools in iCFM, CETA's cloud-based production management platform for VFX.Wednesday, February 5th, 2025Tutorials: Creating a Character for Games - Vol. 1 and Vol. 2Master industry-standard workflows for sculpting and texturing AAA game characters with The Gnomon Workshop's two-part tutorial series.Tuesday, February 4th, 2025More NewsLightWave Digital previews LightWave 2025Check out the new features in the ForestPack 9.1 updatesSee five great VFX breakdowns from the 2025 V-Ray showreelHow Vine FX created a parasitic underwater forest for The RigBoris FX releases Optics 2025Thinkinetic releases Pulldownit 6 for MayaAdobe launches unmetered access to Substance 3D AssetsAdobe releases Photoshop 26.3Check out free Android and iOS 3D scanning app RealityScan 1.6Free tool: Animation Render Time Estimator for BlenderChaos releases V-Ray 7 for HoudiniChaos releases V-Ray 7 for MayaOlder Posts

A three-step process for finding truth onlinePublished inThe Medium BlogSent as aNewsletter3 min readJust now-- Were back!Issue #264: confessions of a healthcare comms lead, executive orders, and brown noiseBy Harris SockelJon Krakauer, bestselling author of several blockbuster nonfiction books including Into the Wild, took to Medium last week to respond to a YouTuber who claims there are significant factual errors in his 1997 book about an Everest climbing disaster, Into Thin Air.This is kind of personal for me. Into Thin Air is not only one of my favorite books, its one of the first books of narrative nonfiction I ever read. I was assigned it as summer reading before 11th grade, after it was named a finalist for the Pulitzer Prize. The book chronicles Krakauers experience climbing Everest in 96 and a deadly storm that killed eight people and left several others stranded on the mountain. Its riveting. Its nonfiction, but its just as engrossing as the best novel youve ever read.Most of the accusations outlined in the post concern the position of climbers on the mountain and how prepared or unprepared the crew was for the storm. I wont get into too much detail, but I do want to briefly touch on the tactic this YouTuber is taking (as alleged by Krakauer) because it points toward a universal law of the internet.Theres an axiom about online discourse known as the Bullshit Asymmetry Principle, Krakauer writes. The less vulgar term for it is Brandolinis Law, named after an Italian programmer who postulated that the energy it takes to refute a single falsehood is an order of magnitude larger than that needed to produce it.In other words, distortion happens fast. Getting to the truth is a slower process.Mike Caulfield, a who studies the spread of rumors and misinformation for a living, has a three-step process for finding the truth on the internet, and I think back to it often whenever Im confronted with (a) someone who seems overconfident about questioning an established fact, or (b) anything that feels out of my depth:Check for previous fact-checking workGo upstream to the primary sourceRead laterally, which basically means Google your sources (you can do this easily with a minus search term, e.g. if you want background on the Wall Street Journal but dont want a bunch of results from the site itself, search: -site:wsj.com Wall Street Journal)Most importantly: Give the truth some time! Krakauers post is the first of what will be eight separate chapters of commentary and (as a huge fan of this book for going on 20 years) Ill be following closely. Were also readingKathleen Murphy, former head of comms for her states BlueCross BlueShield affiliate, recalls arguing with the companys Chief Medical Officer over a denied claim to a child who needed lifetime IV nutrition therapy. Murphy had no choice but to issue a bland statement explaining in deliberately arcane language why the claim was denied, but as she sent it she was shaken. I wanted to scream, she writes, but I had a job to do. So I opened my laptop and composed an email.FDR issued 3,721 executive orders over his 12 years in office, the record by far (second is Woodrow Wilson at 1,803). Most of FDRs orders sought to remake the federal administration but what he did not do, and whats unique to Trumps approach, is use an EO to create and empower a small team of individuals to enact sweeping and substantive reform on the federal civil service, without the consent of Congress. (Peter Stanley Federman, Assistant Prof. of Policy and Global Studies) A top highlight on Medium from last weekWe can disagree and still love each other unless your disagreement is rooted in my oppression. (Tee Mugayi)Your daily dose of practical wisdomIf you (like me) work from home and cant bear the silence but are too distracted by music try white noise? Or brown noise, featuring lower, more calming frequencies a la rushing tides.

Square Enix revenues fall over the past nine months to $1.6bnHD games segment operating profits jump to $30.2 million due to strong sales of Dragon Quest 3 HD-2DImage credit: Square Enix News by Sophie McEvoy Staff Writer Published on Feb. 10, 2025 Square Enix released its financial earnings for the nine months ending December 31, 2024, reporting a decrease in net sales attributed to "weakness in existing titles" in its mobile and PC browser sub-segment.The firm also saw a drop in Digital Entertainments revenue, including its video games business, due to new titles generating lower sales than Final Fantasy 16 and Dragon Quest: The Dark Prince the previous year.Here's what you need to know:The numbers: Net sales: 248.5 billion ($1.6 billion, down 3.5% year-on-year)Profit: 24.7 billion ($161.9 million, down 7.7% year-on-year)Digital entertainment net sales: 160.3 billion ($1.05 billion, down 10.7% year-on-year)The highlights:Despite a drop in overall net sales in its Digital Entertainment segment, Square Enix saw a 70.3% increase in operating profit for its HD games subsection to 4.6 billion ($30.2 million) compared to the same period last year.The firm attributed this to "lower development cost amortisation and advertising expenses", in addition to stronger sales of Dragon Quest 3 HD-2D "than initially assumed" which launched in November 2024.Square Enix also saw a 26.4% rise in net sales for its MMO sub-segment, reaching 44 billion ($288.6 million) following the release of Final Fantasy 14: Dawntrail in July 2024.Looking at its mobile and PC browser sub-segment, the firm saw a significant decline of 58.5% in operating profit to 5.6 billion ($36.7 million) compared to the same period last year.The firm acknowledged that despite the launch of its puzzle RPG Emberstoria in November 2024, there was a drop due to "weakness in existing titles, the recognition of royalty revenue in the previous fiscal year, and a valuation write-down associated with content production account."Looking ahead, Square Enix said there were no changes in its forecast for the fiscal year, which were announced during its FY24 results published last May.

Large Language Models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks, particularly in mathematical problem-solving and coding applications. Research has shown a strong correlation between the length of reasoning chains and improved accuracy in problem-solving outcomes. However, they face significant challenges: while extended reasoning processes enhance problem-solving capabilities, they often lead to inefficient solutions. Models tend to generate unnecessarily lengthy reasoning chains even for simple questions that could be solved more directly. This one-size-fits-all approach to reasoning length creates computational inefficiency and reduces the practical utility of these systems in real-world applications.Various methodologies have emerged to enhance LLMs reasoning capabilities, with Chain-of-Thought (CoT) being a foundational approach that improves problem-solving by breaking down reasoning into discrete steps. Building upon CoT, researchers have developed more complex techniques such as extended CoT with additional steps, self-reflection mechanisms, multi-turn reasoning, and multi-agent debate systems. Recent developments have focused on scaling up reasoning length, as demonstrated by models like OpenAI-o1 and DeepSeek-R1. However, they generate extensive reasoning chains regardless of the problems complexity. This inefficient approach increases computational costs and larger carbon footprints.Researchers from Meta AI and The University of Illinois Chicago have proposed an innovative approach to address the inefficiencies in LLM reasoning by developing a system that automatically adjusts reasoning trace lengths based on query complexity. While previous heuristic methods have attempted to improve token efficiency for better accuracy with reduced overhead, this new research takes a reinforcement learning (RL) perspective. Instead of explicitly modeling response lengths or balancing intrinsic and extrinsic rewards, the researchers have developed a grouping methodology, that involves categorizing responses into distinct groups based on their characteristics, creating a comprehensive framework to cover the entire response space while maintaining efficiency.The proposed methodology employs a sequence-level notation system that simplifies the complex transition probabilities and intermediate rewards by treating each response as a complete unit. The architecture divides responses into two primary groups, one for regular-length Chain-of-Thought responses and the other for extended responses, each with distinct inference costs. The system operates through a bi-level optimization framework, where resource allocation constraints are defined within a convex polytope that limits the density mass of each group. Moreover, the algorithm uses an iterative approach, solving the upper-level problem through gradient updates while directly addressing the lower-level optimization at each iteration.The experimental results demonstrate significant performance improvements across different implementations of the proposed methodology. The supervised fine-tuning (SFT) constructions, SVSFT and ASV-SFT-1, achieve enhanced pass@1 metrics, though at the cost of increased inference requirements. More notably, the ASV-IuB-q+ formulation with parameters set at 50% and 75% show remarkable efficiency improvements, reducing costs by 4.14% at 2.16 times and 5.74% at 4.32 times respectively, matching the performance of SCoRe, a leading RL-based self-correction method. The findings also reveal a noteworthy limitation of prompting-based and SFT-based methods in both absolute improvement and efficiency metrics, suggesting that self-correction capabilities emerge more effectively through RL.In conclusion, researchers introduced a method to overcome the inefficiencies in LLM reasoning. Moreover, they introduced IBPO, a constrained policy optimization framework that implements a weighted Supervised Fine-Tuning update mechanism. This approach determines optimal weights through an integer linear programming solution, in each iteration, built upon the CGPO framework. While the system shows effective constraint adherence and dynamic inference budget allocation in mathematical reasoning tasks, computational resource limitations can be addressed through sample accumulation across multiple steps. Future research directions include expanding the frameworks applicability across different LLM applications and scaling up experimental implementations to test its full potential in various contexts.Check outthePaper.All credit for this research goes to the researchers of this project. Also,dont forget to follow us onTwitterand join ourTelegram ChannelandLinkedIn Group. Dont Forget to join our75k+ ML SubReddit. Sajjad AnsariSajjad Ansari is a final year undergraduate from IIT Kharagpur. As a Tech enthusiast, he delves into the practical applications of AI with a focus on understanding the impact of AI technologies and their real-world implications. He aims to articulate complex AI concepts in a clear and accessible manner.Sajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/Microsoft AI Researchers Release LLaVA-Rad: A Lightweight Open-Source Foundation Model for Advanced Clinical Radiology Report GenerationSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/ACECODER: Enhancing Code Generation Models Through Automated Test Case Synthesis and Reinforcement LearningSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/s1: A Simple Yet Powerful Test-Time Scaling Approach for LLMsSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/Neural SpaceTimes (NSTs): A Class of Trainable Deep Learning-based Geometries that can Universally Represent Nodes in Weighted Directed Acyclic Graphs (DAGs) as Events in a Spacetime Manifold [Recommended] Join Our Telegram Channel