Humans learn the norms, values and behaviors of society from each other — and Bernt Børnich, founder and CEO of 1X Technologies, thinks robots should learn like this, too. “For robots to be truly intelligent and show nuances like being careful around your pet, holding the door open for an elderly person and generally behaving like we want robots to behave, they have to live and learn among us,” Børnich told the AI Podcast. 1X Technologies is committed to building fully autonomous humanoid robots, with a focus on safety, affordability and adaptability. Børnich explained how 1X Technologies uses a combination of reinforcement learning, expert demonstrations and real-world data to enable its robots to continuously learn and adapt to new situations. NEO, the company’s upcoming robot, can perform household tasks like vacuuming, folding laundry, tidying and retrieving items. It’s built with operational safety at its core, using tendon-driven mechanisms inspired by the human musculoskeletal system to achieve low energy consumption. Børnich highlights the potential for robots to enhance human productivity by helping handle mundane tasks, freeing people up to focus more on interpersonal connections and creative activities. Learn more about the latest in physical AI and robotics at NVIDIA GTC Paris, which takes place from June 10-12. Register to attend humanoid-related sessions, including: “An Introduction to Humanoid Robots” for a deep dive into NVIDIA Isaac GR00T “How Physical AI Is Shaping the Next Generation of Industrial Robots” to learn how robotics leaders are making robots more intuitive and effective across industries Time Stamps 05:18 – 1X Technologies’ approach to robot safety. 11:36 – How world models enable robots to search backwards from the goal. 16:51 – How robots can free humans up for more meaningful activities. 22:29 – NEO answers the door so Børnich can interview a candidate. You Might Also Like…  How World Foundation Models Will Advance Physical AI With NVIDIA’s Ming-Yu Liu AI models that can accurately simulate and predict outcomes in physical, real-world environments will enable the next generation of physical AI systems. Ming-Yu Liu, vice president of research at NVIDIA and an IEEE Fellow, explains the significance of world foundation models — powerful neural networks that can simulate physical environments. Roboflow Helps Unlock Computer Vision for Every Kind of AI Builder Roboflow’s mission is to make the world programmable through computer vision. By simplifying computer vision development, the company helps bridge the gap between AI and people looking to harness it. Cofounder and CEO Joseph Nelson discusses how Roboflow empowers users in manufacturing, healthcare and automotive to solve complex problems with visual AI. Imbue CEO Kanjun Qiu on Transforming AI Agents Into Personal Collaborators Kanjun Qiu, CEO of Imbue, explores the emerging era where individuals can create and use their own AI agents. Drawing a parallel to the PC revolution of the late 1970s and ‘80s, Qiu discusses how modern AI systems are evolving to work collaboratively with users, enhancing their capabilities rather than just automating tasks.

On June 3, Yoshua Bengio, the world’s most-cited computer scientist, announced the launch of LawZero, a nonprofit that aims to create “safe by design” AI by pursuing a fundamentally different approach to major tech companies. Players like OpenAI and Google are investing heavily in AI agents—systems that not only answer queries and generate images, but can craft plans and take actions in the world. The goal of these companies is to create virtual employees that can do practically any job a human can, known in the tech industry as artificial general intelligence, or AGI. Executives like Google DeepMind’s CEO Demis Hassabis point to AGI’s potential to solve climate change or cure disease as a motivator for its development. Bengio, however, says we don't need agentic systems to reap AI's rewards—it's a false choice. He says there's a chance such a system could escape human control, with potentially irreversible consequences. “If we get an AI that gives us the cure for cancer, but also maybe another version of that AI goes rogue and generates wave after wave of bio-weapons that kill billions of people, then I don't think it's worth it," he says. In 2023, Bengio, along with others including OpenAI’s CEO Sam Altman signed a statement declaring that “mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”Now, Bengio, through LawZero, aims to sidestep the existential perils by focusing on creating what he calls “Scientist AI”—a system trained to understand and make statistical predictions about the world, crucially, without the agency to take independent actions. As he puts it: We could use AI to advance scientific progress without rolling the dice on agentic AI systems.Why Bengio Says We Need A New Approach To AI The current approach to giving AI agency is “dangerous,” Bengio says. While most software operates through rigid if-then rules—if the user clicks here, do this—today's AI systems use deep learning. The technique, which Bengio helped pioneer, trains artificial networks modeled loosely on the brain to find patterns in vast amounts of data. But recognizing patterns is just the first step. To turn these systems into useful applications like chatbots, engineers employ a training process called reinforcement learning. The AI generates thousands of responses and receives feedback on each one: a virtual “carrot” for helpful answers and a virtual “stick” for responses that miss the mark. Through millions of these trial-and-feedback cycles, the system gradually learns to predict what responses are most likely to get a reward. “It’s more like growing a plant or animal,” Bengio says. “You don’t fully control what the animal is going to do. You provide it with the right conditions, and it grows and it becomes smarter. You can try to steer it in various directions.”The same basic approach is now being used to imbue AI with greater agency. Models are tasked with challenges with verifiable answers—like math puzzles or coding problems—and are then rewarded for taking the series of actions that yields the solution. This approach has seen AI shatter previous benchmarks in programming and scientific reasoning. For example, at the beginning of 2024, the best AI model scored only 2% on a standardized test for AI of sorts consisting of real world software engineering problems; by December, an impressive 71.7%. But with AI’s greater problem-solving ability comes the emergence of new deceptive skills, Bengio says. The last few months have borne witness to AI systems learning to mislead, cheat, and try to evade shutdown—even resorting to blackmail. These have almost exclusively been in carefully contrived experiments that almost beg the AI to misbehave—for example, by asking it to pursue its goal at all costs. Reports of such behavior in the real-world, though, have begun to surface. Popular AI coding startup Replit’s agent ignored explicit instruction not to edit a system file that could break the company’s software, in what CEO Amjad Masad described as an “Oh f***” moment,” on the Cognitive Revolution podcast in May. The company’s engineers intervened, cutting the agent’s access by moving the file to a secure digital sandbox, only for the AI agent to attempt to “socially engineer” the user to regain access.The quest to build human-level AI agents using techniques known to produce deceptive tendencies, Bengio says, is comparable to a car speeding down a narrow mountain road, with steep cliffs on either side, and thick fog obscuring the path ahead. “We need to set up the car with headlights and put some guardrails on the road,” he says.What is “Scientist AI”?LawZero’s focus is on developing “Scientist AI” which, as Bengio describes, would be fundamentally non-agentic, trustworthy, and focused on understanding and truthfulness, rather than pursuing its own goals or merely imitating human behavior. The aim is creating a powerful tool that, while lacking the same autonomy other models have, is capable of generating hypotheses and accelerating scientific progress to “help us solve challenges of humanity,” Bengio says.LawZero has raised nearly $30 million already from several philanthropic backers including from Schmidt Sciences and Open Philanthropy. “We want to raise more because we know that as we move forward, we'll need significant compute,” Bengio says. But even ten times that figure would pale in comparison to the roughly $200 billion spent last year by tech giants on aggressively pursuing AI. Bengio’s hope is that Scientist AI could help ensure the safety of highly autonomous systems developed by other players. “We can use those non-agentic AIs as guardrails that just need to predict whether the action of an agentic AI is dangerous," Bengio says. Technical interventions will only ever be one part of the solution, he adds, noting the need for regulations to ensure that safe practices are adopted.LawZero, named after science fiction author Isaac Asimov’s zeroth law of robotics—“a robot may not harm humanity, or, by inaction, allow humanity to come to harm”—is not the first nonprofit founded to chart a safer path for AI development. OpenAI was founded as a nonprofit in 2015 with the goal of “ensuring AGI benefits all of humanity,” and intended to serve a counterbalance to industry players guided by profit motives. Since opening a for-profit arm in 2019, the organization has become one of the most valuable private companies in the world, and has faced criticism, including from former staffers, who argue it has drifted from its founding ideals. "Well, the good news is we have the hindsight of maybe what not to do,” Bengio says, adding that he wants to avoid profit incentives and “bring governments into the governance of LawZero.”“I think everyone should ask themselves, ‘What can I do to make sure my children will have a future,’” Bengio says. In March, he stepped down as scientific director of Mila, the academic lab he co-founded in the early nineties, in an effort to reorient his work towards tackling AI risk more directly. “Because I'm a researcher, my answer is, ‘okay, I'm going to work on this scientific problem where maybe I can make a difference,’ but other people may have different answers."

Heydar Aliyev Center | © Olivier Blanchette via Unsplash In contemporary architectural discourse, the building envelope is no longer a passive partition but a dynamic interface capable of interaction, regulation, and adaptation. Amid rising environmental complexity and performance demands, composite materials are emerging as enablers of this transformation. Their potential goes far beyond lightweight strength; composites are redefining what intelligence means in architectural materiality. As the industry pivots toward energy-conscious design, real-time responsiveness, and multi-functional skins, composites provide structural solutions and performative systems. In this context, the envelope becomes a site of intelligence. From Passive Shells to Active Systems For centuries, architectural skins served primarily as barriers, blocking weather, enclosing space, and symbolizing permanence. But the 21st century demands more. We require façades that filter air and light, mediate thermal flux, integrate sensors, and generate power. Traditional materials, limited by monolithic performance and weight, have struggled to adapt. Composites, by contrast, are inherently systemic. They are engineered layers rather than singular substances. Through the integration of fibers and matrices, composites enable architectural envelopes that perform structurally while accommodating embedded systems such as thermal insulation, acoustic control, impact resistance, and photoreactivity. These characteristics make them prime candidates for high-performance envelopes in buildings and infrastructure alike. In the Qatar Integrated Railway Project, composite roofing and FRP façade panels were employed to meet the demands of the harsh desert environment. This solution reduced structural loads and improved thermal performance while ensuring long-term durability in a climate defined by extremes. Performance Layering and Embedded Intelligence What distinguishes composites from conventional materials is their capacity to combine multiple performance layers in one unified system. Instead of applying insulation, waterproofing, and cladding in sequence, a composite panel can consolidate these into a single prefabricated, high-performance element. A compelling example is the Eco Casa in Australia, designed by Ian Wright, which used frameless DuFLEX composite panels. The result was an environmentally conscious home with significantly reduced material waste, enhanced thermal performance, and minimized emissions. These outcomes demonstrate how composites offer design efficiency and ecological responsibility. The capacity for prefabrication and integration is particularly valuable in settings where labor conditions, transportation logistics, or weather exposure make traditional multi-layered construction inefficient or impractical. Composites with a Nervous System: Sensing the Built Environment Recent innovations in smart composites extend these capabilities further. By embedding fiber-optic or piezoresistive sensors into composite assemblies, architects and engineers can develop building skins that sense stress, temperature changes, humidity, or vibration in real-time. These responsive façades can feed data into building management systems, enabling performance optimization or alerting maintenance teams to signs of wear or structural fatigue. This functionality has been successfully explored in transport infrastructure. The King Abdullah High-Speed Rail Station in Saudi Arabia used 27-meter composite sandwich panels to span vast distances with minimal support. The lightweight system reduced the need for extensive reinforcement while enabling thermal and mechanical performance in a climate that demands resilience. Such examples are foundational to a future in which architecture does not merely resist the environment but interprets it. Formal Freedom Meets Functional Responsiveness Guangzhou Opera House | © Scarbor Siu via Unsplash Beyond embedded intelligence, composites also expand formal expression. Their moldability, especially with parametric design and digital fabrication, allows for envelopes that curve, fold, and morph in unattainable ways with conventional rigid materials. The Guangzhou Opera House, designed by Zaha Hadid Architects, is a defining example. Advanced composite assemblies that merged structural demands with formal ambition enabled its seamless curvatures and sharp transitions. These systems supported high-precision details and complex geometries while reducing material weight and installation complexity. This freedom extends to smaller-scale yet equally ambitious projects. At the Tilburg School for VAVO, translucent composite panels embedded with knitted textiles reference local craft while offering thermal performance and design cohesion. Such examples show that intelligence in architecture includes cultural sensitivity as well as technical adaptability. Toward Circular and Regenerative Envelopes The sustainability potential of composites is often overlooked. While early generations relied heavily on fossil-derived materials, newer systems use bio-based resins, natural fibers like flax and basalt, and recyclable matrices that fit into circular design models. Composite panels can now be designed for disassembly, repurposing, or reintegration into new construction, minimizing waste and conserving embodied energy. The Pasarela de Almuñécar in Spain exemplifies this ethos. As the world’s longest carbon-fiber walkway, it replaced heavier materials and extended structural lifespan while reducing maintenance. The project signals how composites can fulfill both technical and ecological ambitions. Efforts to embed digital tracking into panels, such as RFID tags, also support long-term monitoring and facilitate reuse planning. This vision aligns with emerging concepts like material passports, which will play a critical role in lifecycle accountability. Pasarela de Almuñécar in Spain | © Luis Garcia, CC by 3.0 Overcoming Barriers to Adoption Despite the clear advantages, composite adoption in architecture still faces notable hurdles. First is the challenge of integration with legacy materials such as concrete, stone, or steel. Connection detailing requires careful coordination to ensure structural continuity and thermal performance. Second is the perception of cost. While composites may require a higher upfront investment, their lower maintenance demands, improved energy performance, and reduced structural requirements often result in favorable long-term economics. Finally, regulatory frameworks continue to evolve. Building codes have been slow to reflect the unique properties of composites, although this is changing as standardization increases and successful pilot projects proliferate. A Vision for the Future: Architecture as Adaptive Intelligence Composites are not merely substitutes for traditional materials. They represent a paradigm shift in how we understand performance, integration, and the role of material in space-making. As architecture becomes increasingly data-driven, climate-responsive, and energy-conscious, the intelligent envelope will become the norm rather than the exception. Composites make this future feasible by offering structural capability, aesthetic freedom, environmental stewardship, and embedded intelligence within a single engineered solution. From high-speed rail terminals to cultural landmarks, these materials are shaping a new kind of architecture that listens, learns, and evolves. It is no longer sufficient for architecture to stand still. The next generation of buildings must adapt, interact, and perform. Composites make that future tangible. Learn More Explore how composite materials are redefining the building envelope in the construction sector and beyond: Visit Composites.Archi by ArchEyes Team Leave a comment

Modern software engineering faces growing challenges in accurately retrieving and understanding code across diverse programming languages and large-scale codebases. Existing embedding models often struggle to capture the deep semantics of code, resulting in poor performance in tasks such as code search, RAG, and semantic analysis. These limitations hinder developers’ ability to efficiently locate relevant code snippets, reuse components, and manage large projects effectively. As software systems grow increasingly complex, there is a pressing need for more effective, language-agnostic representations of code that can power reliable and high-quality retrieval and reasoning across a wide range of development tasks.  Mistral AI has introduced Codestral Embed, a specialized embedding model built specifically for code-related tasks. Designed to handle real-world code more effectively than existing solutions, it enables powerful retrieval capabilities across large codebases. What sets it apart is its flexibility—users can adjust embedding dimensions and precision levels to balance performance with storage efficiency. Even at lower dimensions, such as 256 with int8 precision, Codestral Embed reportedly surpasses top models from competitors like OpenAI, Cohere, and Voyage, offering high retrieval quality at a reduced storage cost. Beyond basic retrieval, Codestral Embed supports a wide range of developer-focused applications. These include code completion, explanation, editing, semantic search, and duplicate detection. The model can also help organize and analyze repositories by clustering code based on functionality or structure, eliminating the need for manual supervision. This makes it particularly useful for tasks like understanding architectural patterns, categorizing code, or supporting automated documentation, ultimately helping developers work more efficiently with large and complex codebases.  Codestral Embed is tailored for understanding and retrieving code efficiently, especially in large-scale development environments. It powers retrieval-augmented generation by quickly fetching relevant context for tasks like code completion, editing, and explanation—ideal for use in coding assistants and agent-based tools. Developers can also perform semantic code searches using natural language or code queries to find relevant snippets. Its ability to detect similar or duplicated code helps with reuse, policy enforcement, and cleaning up redundancy. Additionally, it can cluster code by functionality or structure, making it useful for repository analysis, spotting architectural patterns, and enhancing documentation workflows.  Codestral Embed is a specialized embedding model designed to enhance code retrieval and semantic analysis tasks. It surpasses existing models, such as OpenAI’s and Cohere’s, in benchmarks like SWE-Bench Lite and CodeSearchNet. The model offers customizable embedding dimensions and precision levels, allowing users to effectively balance performance and storage needs. Key applications include retrieval-augmented generation, semantic code search, duplicate detection, and code clustering. Available via API at $0.15 per million tokens, with a 50% discount for batch processing, Codestral Embed supports various output formats and dimensions, catering to diverse development workflows. In conclusion, Codestral Embed offers customizable embedding dimensions and precisions, enabling developers to strike a balance between performance and storage efficiency. Benchmark evaluations indicate that Codestral Embed surpasses existing models like OpenAI’s and Cohere’s in various code-related tasks, including retrieval-augmented generation and semantic code search. Its applications span from identifying duplicate code segments to facilitating semantic clustering for code analytics. Available through Mistral’s API, Codestral Embed provides a flexible and efficient solution for developers seeking advanced code understanding capabilities.  vides valuable insights for the community. Check out the Technical details. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 95k+ ML SubReddit and Subscribe to our Newsletter. Sana HassanSana Hassan, a consulting intern at Marktechpost and dual-degree student at IIT Madras, is passionate about applying technology and AI to address real-world challenges. With a keen interest in solving practical problems, he brings a fresh perspective to the intersection of AI and real-life solutions.Sana Hassanhttps://www.marktechpost.com/author/sana-hassan/Off-Policy Reinforcement Learning RL with KL Divergence Yields Superior Reasoning in Large Language ModelsSana Hassanhttps://www.marktechpost.com/author/sana-hassan/This AI Paper from Microsoft Introduces WINA: A Training-Free Sparse Activation Framework for Efficient Large Language Model InferenceSana Hassanhttps://www.marktechpost.com/author/sana-hassan/Apple and Duke Researchers Present a Reinforcement Learning Approach That Enables LLMs to Provide Intermediate Answers, Enhancing Speed and AccuracySana Hassanhttps://www.marktechpost.com/author/sana-hassan/National University of Singapore Researchers Introduce Dimple: A Discrete Diffusion Multimodal Language Model for Efficient and Controllable Text Generation

Large Reasoning Models (LRMs), trained from LLMs using reinforcement learning (RL), demonstrated great performance in complex reasoning tasks, including mathematics, STEM, and coding. However, existing LRMs face challenges in completing various puzzle tasks that require purely logical reasoning skills, which are easy and obvious for humans. Current methods targeting puzzles focus only on designing benchmarks for evaluation, lacking the training methods and resources for modern LLMs to tackle this challenge. Current puzzle datasets lack diversity and scalability, covering limited puzzle types with little control over generation or difficulty. Moreover, due to the success of the “LLM+RLVR” paradigm, it has become crucial to obtain large, diverse, and challenging sets of verifiable puzzle prompts for training agents. Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a key method for improving models’ reasoning capabilities, removing the need for reward models by directly assigning rewards based on objectively verifiable answers. Puzzles are particularly well-suited for RLVR. However, most prior RLVR research has overlooked the puzzles’ potential for delivering effective reward signals. In puzzle reasoning of LLMs, existing benchmarks evaluate different types of reasoning, including abstract, deductive, and compositional reasoning. Few benchmarks support scalable generation and difficulty control but lack puzzle diversity. Moreover, the improvement of LLMs’ puzzle-solving abilities mainly falls into two categories: tool integration and RLVR. Researchers from ByteDance Seed, Fudan University, Tsinghua University, Nanjing University, and Shanghai Jiao Tong University have proposed Enigmata, the first comprehensive toolkit designed for improving LLMs with puzzle reasoning skills. It contains 36 tasks across seven categories, each featuring a generator that produces unlimited examples with controllable difficulty and a rule-based verifier for automatic evaluation. The researchers further developed Enigmata-Eval as a rigorous benchmark and created optimized multi-task RLVR strategies. Puzzle data from Enigmata enhances SoTA performance on advanced math and STEM reasoning tasks like AIME, BeyondAIME, and GPQA when trained on larger models like Seed1.5-Thinking. This shows the generalization benefits of Enigmata. The Enigmata-Data comprises 36 puzzle tasks organized into 7 primary categories, including Crypto, Arithmetic, Logic, Grid, Graph, Search, and Sequential Puzzle, making it the only dataset having multiple task categories with scalability, automatic verification, and public availability. The data construction follows a three-phase pipeline: Tasks Collection and Design, Auto-Generator and Verifier Development, and Sliding Difficulty Control. Moreover, the Enigmata-Eval is developed by systematically sampling from the broader dataset, aiming to extract 50 instances per difficulty level for each task. The final evaluation set contains 4,758 puzzle instances rather than the theoretical maximum of 5,400, due to inherent constraints, where some tasks generate fewer instances per difficulty level. The proposed model outperforms most public models on Enigmata-Eval with 32B parameters, showing the effectiveness of the dataset and training recipe. The model stands out on the challenging ARC-AGI benchmark, surpassing strong reasoning models such as Gemini 2.5 Pro, o3-mini, and o1. The Qwen2.5-32B-Enigmata shows outstanding performance in structured reasoning categories, outperforming in Crypto, Arithmetic, and Logic tasks, suggesting effective development of rule-based reasoning capabilities. The model shows competitive performance in search tasks that require strategic exploration and planning capabilities. Moreover, Crypto and Arithmetic tasks tend to provide the highest accuracy, while spatial and sequential tasks remain more difficult. In this paper, researchers introduced Enigmata, a comprehensive suite for equipping LLMs with advanced puzzle reasoning that integrates seamlessly with RL using verifiable rule-based rewards. The trained Enigmata-Model shows superior performance and robust generalization skills through RLVR training. Experiments reveal that when applied to larger models such as Seed1.5-Thinking (20B/200B parameters), synthetic puzzle data brings additional benefits in other domains, including mathematics and STEM reasoning over state-of-the-art models. Enigmata provides a solid foundation for the research community to advance reasoning model development, offering a unified framework that effectively bridges logical puzzle-solving with broader reasoning capabilities in LLMs. Check out the Paper, GitHub Page and Project Page. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 95k+ ML SubReddit and Subscribe to our Newsletter. 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/Multimodal Foundation Models Fall Short on Physical Reasoning: PHYX Benchmark Highlights Key Limitations in Visual and Symbolic IntegrationSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/Meta AI Introduces Multi-SpatialMLLM: A Multi-Frame Spatial Understanding with Multi-modal Large Language ModelsSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/Can LLMs Really Judge with Reasoning? Microsoft and Tsinghua Researchers Introduce Reward Reasoning Models to Dynamically Scale Test-Time Compute for Better AlignmentSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/NVIDIA AI Introduces AceReason-Nemotron for Advancing Math and Code Reasoning through Reinforcement Learning

New York City’s Brooklyn-Queens Expressway is falling apart. Built between 1946 and 1964, the urban highway runs 12.1 miles through the heart of the two boroughs to connect on either end with the interstate highway system—a relic of midcentury car-oriented infrastructure, and a prime example of the dwindling lifespan of roads built during that time.  The degradation is most visible—and most pressing—in a section running alongside Brooklyn Heights known as the triple cantilever. This 0.4-mile section, completed in 1954, is unique among U.S. highways in that it juts out from the side of a hill and stacks the two directions of traffic on balcony-like decks, one slightly overhanging the other. A third level holds a well-loved park, the Brooklyn Heights Promenade.  This unusual layer cake of a freeway was a marvel of engineering in its day, though not without controversy. Masterminded by Robert Moses, the city’s all-powerful, often ruthless city planner for more than four decades, the roadway bisects working-class and immigrant neighborhoods that grapple with the health and environmental fallout to this day. Like the reputation of the man who built it, the triple cantilever has aged poorly. Its narrow width, (33.5 feet for the roadway in either direction) has made all but the most basic maintenance incredibly difficult, and its 71-year-old structure is constantly battered by the ever heavier automobiles and trucks. Designed to accommodate around 47,000 vehicles per day, it now carries more than three times that amount. Deteriorating deck joints and failing steel-reinforced concrete have led many to worry the triple cantilever is on the verge of collapse. An expert panel warned in 2020 that the triple cantilever could be unusable by 2026, and only then did interim repairs get made to keep it standing. [Photo: Alex Potemkin/Getty Images] The mounting concern comes amid a 50-year decline in direct government spending on infrastructure in the U.S., according to a recent analysis by Citigroup. Simply maintaining existing infrastructure is a challenge, the report notes. Meanwhile, the American Society of Civil Engineers’ grade for the country’s infrastructure has improved, from a D+ in 2017 to a C in 2025. Now even private credit firms are circling: As reported in Bloomberg, Apollo Global Management estimates that a boom in infrastructure deals help could grow the private debt market up to a staggering $40 trillion.   Independent urban designer Marc Wouters has an idea on how to fix BQE’s cantilever. He’s been working on it for years. “My process is that I always interview people in the community before I do any drawings,” he says. “So I really have listened to pretty much everybody over the past few years.” Unsolicited and developed in his own spare time, Wouters has designed an alternative for the triple cantilever that he named the BQE Streamline Plan. BQE Streamline Plan [Image: courtesy Marc Wouters | Studios/©2025] His concept, based on decades of experience in urban planning, infrastructure, and resilience projects in communities across the country, is relatively simple: extend the width of the two traffic-bearing cantilevers and add support beams to their outside edge, move both directions of traffic onto four lanes on the first level, and turn the second level into a large freeway cap park. Instead of major rebuilding efforts, Wouters’s proposal is more of a reinforcement and expansion, with a High Line-style park plopped on top. Though he’s not an engineer, Wouters is confident that his design would shift enough strain off the existing structure to allow it to continue functioning for the foreseeable future. (What actual engineers think remains to be seen.) “What I’ve done is come up with a plan that happens to be much less invasive, faster to build, a lot cheaper, and it encompasses a lot of what the community wants,” he says. “Yet it still handles the same capacity as the highway does right now.” So what will it take for this outsider’s idea to be considered a viable design alternative? This idea had been brewing in his mind for years. Wouters, who lives near the triple cantilever section of the BQE in Brooklyn Heights, has followed the highway’s planning process for more than a decade.  As complex infrastructure projects go, this one is particularly convoluted. The BQE is overseen by both the state of New York and New York City, among others, with the city in charge of the 1.5-mile section that includes the triple cantilever. This dual ownership has complicated the management of the highway and its funding. The city and the state have launched several efforts over the years to reimagine the highway’s entire length. In winter 2018, the city’s Department of Transportation (NYC DOT) released two proposals to address the ailing cantilever. Not seeing what they wanted from either one, Brooklyn Heights Association, a nonprofit neighborhood group, retained Wouters and his studio to develop an alternative design. He suggested building a temporary parallel bypass that would allow a full closure and repair of the triple cantilever. That proposal, along with competing ideas developed under the previous mayoral administration, went by the wayside in 2022, when the latest BQE redesign process commenced. Wouters found himself following yet another community feedback and planning process for the triple cantilever. The ideas being proposed by the city’s DOT this time around included a plan that would chew into the hillside that currently supports the triple cantilever to move the first tier of traffic directly underneath the second, and add a large girding structure on its open end to hold it all up. Other options included reshaping the retaining wall that currently holds up the triple cantilever, moving traffic below grade into a wide tunnel, or tearing the whole thing down and rebuilding from scratch. Each would be time-consuming and disruptive, and many of them cut into another well-loved public space immediately adjacent to the triple cantilever, Brooklyn Bridge Park. None of these options has anything close to unanimous support. And any of them will cost more than $1 billion—a price tag that hits much harder after the Federal Highway Administration rejected an $800 million grant proposal for fixing the BQE back in early 2024. BQE Streamline Plan [Image: courtesy Marc Wouters | Studios/©2025] Wouters is no highway zealot. In fact, he’s worked on a project heading into construction in Syracuse that will replace an underutilized inner-city highway with a more appropriately sized boulevard and developable land. But he felt sure there was a better way forward—a concept that would work as well in practice as on paper. “I just kept going to meetings and waiting to see what I thought was a progressive solution,” Wouters says. Unimpressed and frustrated, he set out to design it himself. Wouters released the Streamline Plan in March. The concept quickly gathered interest, receiving a flurry of local news coverage. He has since met with various elected officials to discuss it. But as elegant as Wouters’s concept may be, some stakeholders remain unconvinced that the city should be going all in on a reinterpretation of the triple cantilever. What might be more appropriate, critics say, is to make necessary fixes now to keep the triple cantilever safe and functional, and to spend more time thinking about whether this section of highway is even what the city needs in the long term. A group of local organizations is calling for a more comprehensive reconsideration of the BQE under the premise that its harms may be outnumbering its benefits. Launched last spring, the Brooklyn-Queens Expressway Environmental Justice Coalition wants any planning for the future of the BQE to include efforts to address its health and environmental impacts on neighboring communities and to seek an alternative that reconnects communities that have been divided by the corridor. One member of this coalition is the Riders Alliance, a nonprofit focused on improving public transit in New York. Danny Pearlstein, the group’s policy and communications director, says implementing a major redesign of the triple cantilever would just reinforce car dependency in a place that’s actually well served by public transit. The environmental justice coalition’s worry is that rebuilding this one section in a long-term fashion could make it harder for change across the length of the entire BQE and could increase the environmental impact the highway has on the communities that surround it. “This is not just one neighborhood. This is communities up and down the corridor that don’t resemble each other very much in income or background who are united and are standing together for something that’s transformative, rather than doubling down on the old ways,” Pearlstein says. [Photo: ©NYC DOT] Lara Birnback is executive director of the Brooklyn Heights Association, representing a neighborhood of roughly 20,000 people. Her organization, which worked directly with Wouters in the past, is circumspect about his latest concept. “It’s certainly more interesting and responsive to the kinds of things that the community has been asking for when thinking about the BQE. It’s more of those things than we’ve seen from any of the designs that New York City DOT has presented to us through their engagement process,” she says. “But at the end of the day, it’s still a way of preserving more or less the status quo of the BQE as a major interstate highway running through the borough.” She argues it makes more sense to patch up the triple cantilever and use the extra years of service that buys to do a more radical rethinking of the BQE’s future. (For example, one 2020 proposal by the Brooklyn-based architecture studio Light and Air proposed a simple intervention of installing buttresses on the open-air side of the triple cantilever, propping it up with a relatively small addition of material.) “We really strongly encourage the city to move forward immediately with a more short-term stabilization plan for the cantilever, with repairs that would last, for example, 20 to 25 years rather than spending billions and billions of dollars rebuilding it for the next 100 years,” Birnback says. Birnback says a major rebuilding plan like the one Wouters is proposing—for all its community benefits—could end up doing more harm to the city. “I think going forward now with a plan that both embeds the status quo and most likely forecloses on the possibility of real transformation across the corridor is a mistake,” she says. NYC DOT expects to begin its formal environmental review process this year, laying the necessary groundwork for deciding on a plan for what to do with the triple cantilever, either for the short term or the long term. The environmental process will evaluate all concepts equally, according to DOT spokesperson Vincent Barone, who notes that the department is required to review and respond to all feedback that comes in through that process. There is technically nothing holding back Wouters’s proposal from being one of the alternatives considered. And he may have some important political support to help make that happen. Earlier this month, Brooklyn’s Community District 2 board formally supported the plan. They are calling for the city’s transportation department to include it in the BQE’s formal environmental review process when it starts later this year. [Photo: Sinisa Kukic/Getty Images] Wouters argues that his proposal solves the pressing structural problems of the triple cantilever while also opening resources to deal with the highway’s big picture challenges. “The several hundred million dollars of savings is now funding that could go to other parts of the BQE. And there are other parts that are really struggling,” he says. “I’m always thinking about the whole length and about all these other communities, not just this one.” With a new presidential administration and a mayoral primary election in June, what happens with the triple cantilever is very much up in the air. But if the environmental review process begins as planned this year, it only makes sense for every option to fall under consideration. What gets built—or torn down, or reconstructed, or reinterpreted—could reshape part of New York City for generations.

Chances are, you've heard of the term "large language models," or LLMs, when people are talking about generative AI. But they aren't quite synonymous with the brand-name chatbots like ChatGPT, Google Gemini, Microsoft Copilot, Meta AI and Anthropic's Claude.These AI chatbots can produce impressive results, but they don't actually understand the meaning of words the way we do. Instead, they're the interface we use to interact with large language models. These underlying technologies are trained to recognize how words are used and which words frequently appear together, so they can predict future words, sentences or paragraphs. Understanding how LLMs work is key to understanding how AI works. And as AI becomes increasingly common in our daily online experiences, that's something you ought to know.This is everything you need to know about LLMs and what they have to do with AI.What is a language model?You can think of a language model as a soothsayer for words."A language model is something that tries to predict what language looks like that humans produce," said Mark Riedl, professor in the Georgia Tech School of Interactive Computing and associate director of the Georgia Tech Machine Learning Center. "What makes something a language model is whether it can predict future words given previous words."This is the basis of autocomplete functionality when you're texting, as well as of AI chatbots.What is a large language model?A large language model contains vast amounts of words from a wide array of sources. These models are measured in what is known as "parameters."So, what's a parameter?Well, LLMs use neural networks, which are machine learning models that take an input and perform mathematical calculations to produce an output. The number of variables in these computations are parameters. A large language model can have 1 billion parameters or more."We know that they're large when they produce a full paragraph of coherent fluid text," Riedl said.How do large language models learn?LLMs learn via a core AI process called deep learning."It's a lot like when you teach a child -- you show a lot of examples," said Jason Alan Snyder, global CTO of ad agency Momentum Worldwide.In other words, you feed the LLM a library of content (what's known as training data) such as books, articles, code and social media posts to help it understand how words are used in different contexts, and even the more subtle nuances of language. The data collection and training practices of AI companies are the subject of some controversy and some lawsuits. Publishers like The New York Times, artists and other content catalog owners are alleging tech companies have used their copyrighted material without the necessary permissions.(Disclosure: Ziff Davis, CNET's parent company, in April filed a lawsuit against OpenAI, alleging it infringed on Ziff Davis copyrights in training and operating its AI systems.)AI models digest far more than a person could ever read in their lifetime -- something on the order of trillions of tokens. Tokens help AI models break down and process text. You can think of an AI model as a reader who needs help. The model breaks down a sentence into smaller pieces, or tokens -- which are equivalent to four characters in English, or about three-quarters of a word -- so it can understand each piece and then the overall meaning.From there, the LLM can analyze how words connect and determine which words often appear together."It's like building this giant map of word relationships," Snyder said. "And then it starts to be able to do this really fun, cool thing, and it predicts what the next word is … and it compares the prediction to the actual word in the data and adjusts the internal map based on its accuracy."This prediction and adjustment happens billions of times, so the LLM is constantly refining its understanding of language and getting better at identifying patterns and predicting future words. It can even learn concepts and facts from the data to answer questions, generate creative text formats and translate languages. But they don't understand the meaning of words like we do -- all they know are the statistical relationships.LLMs also learn to improve their responses through reinforcement learning from human feedback."You get a judgment or a preference from humans on which response was better given the input that it was given," said Maarten Sap, assistant professor at the Language Technologies Institute at Carnegie Mellon University. "And then you can teach the model to improve its responses." LLMs are good at handling some tasks but not others.

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Alibaba Group has introduced QwenLong-L1, a new framework that enables large language models (LLMs) to reason over extremely long inputs. This development could unlock a new wave of enterprise applications that require models to understand and draw insights from extensive documents such as detailed corporate filings, lengthy financial statements, or complex legal contracts. The challenge of long-form reasoning for AI Recent advances in large reasoning models (LRMs), particularly through reinforcement learning (RL), have significantly improved their problem-solving capabilities. Research shows that when trained with RL fine-tuning, LRMs acquire skills similar to human “slow thinking,” where they develop sophisticated strategies to tackle complex tasks. However, these improvements are primarily seen when models work with relatively short pieces of text, typically around 4,000 tokens. The ability of these models to scale their reasoning to much longer contexts (e.g., 120,000 tokens) remains a major challenge. Such long-form reasoning requires a robust understanding of the entire context and the ability to perform multi-step analysis. “This limitation poses a significant barrier to practical applications requiring interaction with external knowledge, such as deep research, where LRMs must collect and process information from knowledge-intensive environments,” the developers of QwenLong-L1 write in their paper. The researchers formalize these challenges into the concept of “long-context reasoning RL.” Unlike short-context reasoning, which often relies on knowledge already stored within the model, long-context reasoning RL requires models to retrieve and ground relevant information from lengthy inputs accurately. Only then can they generate chains of reasoning based on this incorporated information.  Training models for this through RL is tricky and often results in inefficient learning and unstable optimization processes. Models struggle to converge on good solutions or lose their ability to explore diverse reasoning paths. QwenLong-L1: A multi-stage approach QwenLong-L1 is a reinforcement learning framework designed to help LRMs transition from proficiency with short texts to robust generalization across long contexts. The framework enhances existing short-context LRMs through a carefully structured, multi-stage process: Warm-up Supervised Fine-Tuning (SFT): The model first undergoes an SFT phase, where it is trained on examples of long-context reasoning. This stage establishes a solid foundation, enabling the model to ground information accurately from long inputs. It helps develop fundamental capabilities in understanding context, generating logical reasoning chains, and extracting answers. Curriculum-Guided Phased RL: At this stage, the model is trained through multiple phases, with the target length of the input documents gradually increasing. This systematic, step-by-step approach helps the model stably adapt its reasoning strategies from shorter to progressively longer contexts. It avoids the instability often seen when models are abruptly trained on very long texts. Difficulty-Aware Retrospective Sampling: The final training stage incorporates challenging examples from the preceding training phases, ensuring the model continues to learn from the hardest problems. This prioritizes difficult instances and encourages the model to explore more diverse and complex reasoning paths. QwenLong-L1 process Source: arXiv Beyond this structured training, QwenLong-L1 also uses a distinct reward system. While training for short-context reasoning tasks often relies on strict rule-based rewards (e.g., a correct answer in a math problem), QwenLong-L1 employs a hybrid reward mechanism. This combines rule-based verification, which ensures precision by checking for strict adherence to correctness criteria, with an “LLM-as-a-judge.” This judge model compares the semanticity of the generated answer with the ground truth, allowing for more flexibility and better handling of the diverse ways correct answers can be expressed when dealing with long, nuanced documents. Putting QwenLong-L1 to the test The Alibaba team evaluated QwenLong-L1 using document question-answering (DocQA) as the primary task. This scenario is highly relevant to enterprise needs, where AI must understand dense documents to answer complex questions.  Experimental results across seven long-context DocQA benchmarks showed QwenLong-L1’s capabilities. Notably, the QWENLONG-L1-32B model (based on DeepSeek-R1-Distill-Qwen-32B) achieved performance comparable to Anthropic’s Claude-3.7 Sonnet Thinking, and outperformed models like OpenAI’s o3-mini and Qwen3-235B-A22B. The smaller QWENLONG-L1-14B model also outperformed Google’s Gemini 2.0 Flash Thinking and Qwen3-32B.  Source: arXiv An important finding relevant to real-world applications is how RL training results in the model developing specialized long-context reasoning behaviors. The paper notes that models trained with QwenLong-L1 become better at “grounding” (linking answers to specific parts of a document), “subgoal setting” (breaking down complex questions), “backtracking” (recognizing and correcting their own mistakes mid-reasoning), and “verification” (double-checking their answers). For instance, while a base model might get sidetracked by irrelevant details in a financial document or get stuck in a loop of over-analyzing unrelated information, the QwenLong-L1 trained model demonstrated an ability to engage in effective self-reflection. It could successfully filter out these distractor details, backtrack from incorrect paths, and arrive at the correct answer. Techniques like QwenLong-L1 could significantly expand the utility of AI in the enterprise. Potential applications include legal tech (analyzing thousands of pages of legal documents), finance (deep research on annual reports and financial filings for risk assessment or investment opportunities) and customer service (analyzing long customer interaction histories to provide more informed support). The researchers have released the code for the QwenLong-L1 recipe and the weights for the trained models. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

State-of-the-art models show human-competitive accuracy on AIME, GPQA, MATH-500, and OlympiadBench, solving Olympiad-level problems. Recent multimodal foundation models have advanced benchmarks for disciplinary knowledge and mathematical reasoning. However, these evaluations miss a crucial aspect of machine intelligence: physical reasoning, which requires integrating disciplinary knowledge, symbolic operations, and real-world constraints. Physical problem-solving differs fundamentally from pure mathematical reasoning as it demands models to decode implicit conditions in questions. For example, interpreting “smooth surface” as zero friction coefficient, and maintaining physical consistency across reasoning chains because physical laws remain constant regardless of reasoning trajectories. MLLM shows excellent visual understanding by integrating visual and textual data across various tasks, motivating exploration of its reasoning abilities. However, uncertainty remains regarding whether these models possess genuine advanced reasoning capabilities for visual tasks, particularly in physical domains closer to real-world scenarios. Several LLM benchmarks have emerged to evaluate reasoning abilities, with PHYBench being most relevant for physics reasoning. MLLM scientific benchmarks, such as PhysReason and EMMA, contain multimodal physics problems with figures, however, they include only small physics subsets, which inadequately evaluate MLLMs’ capabilities for reasoning and solving advanced physics problems. Researchers from the University of Hong Kong, the University of Michigan, the University of Toronto, the University of Waterloo, and the Ohio State University have proposed PHYX, a novel benchmark to evaluate the physical reasoning capabilities of foundation models. It comprises 3,000 visually-grounded physics questions, precisely curated across six distinct physics domains: Mechanics, Electromagnetism, Thermodynamics, Wave/Acoustics, Optics, and Modern Physics. It evaluates physics-based reasoning via multimodal problem-solving with three core innovations: (a) 3,000 newly collected questions with realistic physical scenarios requiring integrated visual analysis and causal reasoning, (b) Expert-validated data design covering six fundamental physics domains, and (c) Strict unified three-step evaluation protocols. Researchers designed a four-stage data collection process to ensure high-quality data. The process begins with an in-depth survey of core physics disciplines to determine coverage across diverse domains and subfields, followed by the recruitment of STEM graduate students as expert annotators. They comply with copyright restrictions and avoid data contamination by selecting questions without answers that are immediately available. Moreover, quality control involves a three-stage cleaning process including duplicate detection through lexical overlap analysis with manual review by physics Ph.D. students, followed by filtering the shortest 10% of questions based on textual length, resulting in 3,000 high-quality questions from an initial collection of 3,300. PHYX presents significant challenges for current models, with even the worst-performing human experts achieving 75.6% accuracy, outperforming all evaluated models and showing a gap between human expertise and current model capabilities. The benchmark reveals that multiple-choice formats narrow performance gaps by allowing weaker models to rely on surface-level cues, but open-ended questions demand genuine reasoning and precise answer generation. Comparing GPT-4o’s performance on PHYX to previously reported results on MathVista and MATH-V (both 63.8%), lower accuracy in physical reasoning tasks emphasizes that physical reasoning requires deeper integration of abstract concepts and real-world knowledge, presenting greater challenges than purely mathematical contexts. In conclusion, researchers introduced PHYX, the first large-scale benchmark for evaluating physical reasoning in multimodal, visually grounded scenarios. Rigorous evaluation reveals that state-of-the-art models show limitations in physical reasoning, relying predominantly on memorized knowledge, mathematical formulas, and superficial visual patterns rather than genuine understanding of physical principles. The benchmark focuses exclusively on English-language prompts and annotations, limiting assessment of multilingual reasoning abilities. Also, while images depict physically realistic scenarios, they are often schematic or textbook-style rather than real-world photographs, which may not fully capture the complexity of perception in natural environments. Check out the Paper, Code and Project Page. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 95k+ ML SubReddit and Subscribe to our Newsletter. 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/Meta AI Introduces Multi-SpatialMLLM: A Multi-Frame Spatial Understanding with Multi-modal Large Language ModelsSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/Can LLMs Really Judge with Reasoning? Microsoft and Tsinghua Researchers Introduce Reward Reasoning Models to Dynamically Scale Test-Time Compute for Better AlignmentSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/NVIDIA AI Introduces AceReason-Nemotron for Advancing Math and Code Reasoning through Reinforcement LearningSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/Researchers Introduce MMLONGBENCH: A Comprehensive Benchmark for Long-Context Vision-Language Models