Large Reasoning Models, trained from LLMs using reinforcement learning, 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 Rewardshas 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, 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 #enigmatas #multistage #mixtraining #reinforcement #learning

ZARA Nanjing Xinjiekou | © Seth Powers, ZARA Within the dense commercial fabric of Nanjing’s Xinjiekou district, AIM Architecture’s intervention for ZARA transcends the conventional boundaries of retail design. Rather than presenting a sealed envelope, the store engages directly with the public realm, positioning itself as an architectural node in the city’s pedestrian network. The project acknowledges the evolving role of high-street retail in an era increasingly defined by digital commerce. In this context, the store becomes a civic gesture, serving as a point of sale and an interface between the brand and the city. ZARA Nanjing Xinjiekou Flagship Store Technical Information Architects1-13: AIM Architecture Location: Xinjiekou, Nanjing, China Client: ZARA CommercialCo., Ltd Area: 3,450 m2 | 37,135 Sq. Ft. Project Year: 2023 – 2025 Photographs: © Seth Powers, ZARA We designed the ZARA Nanjing flagship as a seamless extension of the city, a space where retail, public life, and cultural engagement converge to reshape the role of the store in the urban landscape. – AIM Architecture Architects ZARA Nanjing Xinjiekou Flagship Store Photographs © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA © Seth Powers, ZARA Contextual Integration and Urban Engagement A key architectural move is the introduction of a generous cantilevered canopy and fluid facade, softening the threshold between the street and the interior. This transitional zone is supported by a reconfigured streetscape and entrance plaza, inviting occupation beyond the function of shopping. These spatial devices are calibrated to intensify the porosity between inside and outside, fostering a continuity of urban experience. The architectural language, while assertive, avoids spectacle. Instead, it aligns with the rhythms of the street, suggesting a model for how retail environments might more meaningfully participate in the life of the city. AIM Architecture eschews traditional retail hierarchies in favor of a layered and adaptive spatial arrangement. The design is organized around a sequence of interrelated zones, each calibrated to support distinct experiential intensities. At the ground level, the introduction of ZARA Salon offers a refined, intimate encounter with the brand. This contrasts with the upper levels, where spatial arrangements and materiality adopt a more experimental and energetic tone. These programmatic shifts articulate a nuanced understanding of the store’s diverse user profiles and behavioral patterns. Central to the spatial composition is the double-height entrance plaza, conceived as a flexible platform for cultural engagement. Rather than being subordinated to commercial imperatives, this space opens itself to temporary exhibitions, artistic collaborations, and public events. The integration of a visible automated clothing transportation system along a glazed wall serves a dual function: operational efficiency and spatial expression. By revealing the store’s logistical mechanisms, the architecture foregrounds process as a form of engagement, reinforcing the store’s identity as an evolving machine. Material Expression and Architectural Language Material choices throughout the project articulate a sensibility rooted in clarity, tactility, and contextual responsiveness. The palette includes exposed structural concrete, red brick, stainless steel, solid wood, and ultra-clear glass. These materials are not merely applied as surface treatments but are used to express the building’s spatial and structural logic. One of the project’s most distinctive features is the red brick “mountain” staircase that extends toward the public realm. This element performs multiple roles: it serves as circulation, seating, display, and an urban landmark. Its textured materiality references the local architectural vernacular, grounding the otherwise contemporary design in its geographic and cultural context. Inside, the polished concrete floors and cassette ceiling system establishes a framework of disciplined neutrality, allowing the dynamic display environments and customer movement to animate the space without visual clutter. The approach to detailing reflects a commitment to architectural integrity. Junctions are resolved with precision, and structural elements are left deliberately exposed, contributing to a language of authenticity. The lighting strategy, developed in collaboration with Fagerhult, underscores spatial transitions and accentuates material contrasts without resorting to theatricality. Toward a New Retail Paradigm: Cultural Infrastructure The ZARA Nanjing Xinjiekou Flagship Store points toward a redefinition of the retail typology, aligning it more closely with the functions of cultural infrastructure. This is a project where architecture mediates between commerce and community, suggesting that the store can operate as a space of encounter, interaction, and expression. In a moment where physical retail must justify its relevance against the efficiency of e-commerce, AIM Architecture proposes a counter-model in which the store contributes to the public sphere, both spatially and culturally. The project offers an expanded understanding of what a flagship store can be through architectural interventions that encourage lingering discovery and participation. It is a space of consumption and a spatial framework for collective experience. ZARA Nanjing Xinjiekou Flagship Store Plans Ground Level | © AIM Architecture Upper Level | © AIM Architecture ZARA Nanjing Xinjiekou Flagship Store Image Gallery About AIM Architecture AIM Architecture is a Shanghai-based architectural practice founded by Wendy Saunders and Vincent de Graaf. It is known for its multidisciplinary approach that blends architecture, interior design, and urban design. The studio focuses on creating contextually responsive and experientially rich environments that engage with the cultural, social, and spatial dynamics of contemporary cities. Credits and Additional Notes Design Principals: Wendy Saunders, Vincent de Graaf Project Manager: Marta Pozo, Sacha Silva Project Architect: Ewa Szajda Design Team: Alba Galan, Davide Signorato, Emilio Wang, Gabrielle Liu, Guanlin Li, Haochen Yang, Jerry Guo, Song Jie, Kang Jin, Junger Xia, Kexin Gao, Mia Lei, Nora Liu, Victor Mongin, Wei Zheng, Weisha Dai, Qianqian Bao, Xiao Wen, Zhang Yi, Zhao Na Visualization: JIAO Yan Structural Engineer & Façade Consultant: bespoke. Sur-Mesure Engineering Studio General Contractor: Shanghai Choyoin Construction Group Co., Ltd Façade Contractor: Shanghai Choyoin Construction Group Co., Ltd Furniture: Caamaño AsiaFurniture and Complements Co., Ltd  Furniture and Complements Co., Ltd,  Lighting Supplier: Fagerhult Audiovisual Supplier: Inusual Comunicación Innovadora / ICI Technology Shenzhen Ltd. Zacaffè Design: Art Recherche Industrie Digital Experience: Bagel Affairs #zara #nanjing #xinjiekou #flagship #store

The idea of using the body’s own organs as mini bioreactors to grow replacement tissue or even regenerate other organs might sound like something out of a science fiction movie, but it's already becoming reality in cutting-edge labs around the world.A collaboration between Wenzhou Medical University, Nanjing University, and the University of Macau has taken an unexpected turn in regenerative medicine by turning to the spleen, a lymphatic organ typically overshadowed by its more high-profile neighbors. Their findings, recently published in Science Translational Medicine, suggest that the spleen could be key to growing new, functional tissues within the body. And the implications are huge, particularly for diseases like type 1 diabetes.Reinventing the Spleen's PurposeRoughly the size of an avocado and tucked under the left side of the rib cage just above the stomach, the spleen’s usual responsibilities include filtering damaged blood cells and supporting the immune system. It’s often considered non-essential as many people live healthy lives without their spleen, if it was removed after injury or illness.But its seemingly simple structure might be exactly what makes it so powerful. With its sponge-like texture, nutrient-rich environment, and proximity to major blood vessels like those of the liver, the spleen turns out to be an ideal candidate for tissue cultivation.Insulin, Made in the SpleenIn this study, researchers set their sights on type 1 diabetes, a condition in which the immune system destroys insulin-producing pancreatic islet cells. Working with the spleens of primates, the team engineered microenvironments within the test organs to support human pancreatic islets.“We’re essentially converting the spleen into a high-performance bioreactor,” explained study co-author Lei Dong in a press release. “By enhancing extracellular matrix support, accelerating blood vessel growth, and suppressing immune attacks, we’ve created an ideal niche for transplanted cells to thrive.”After transplantation, the human islet cells matured inside the primates’ spleens and began producing insulin and C-peptidecontinuously for 28 days. It’s a critical proof of concept — showing that not only can the spleen host new tissue, but it can support it long enough to function effectively.This isn't the team’s first foray into reimagining the spleen’s capabilities. They already reprogrammed mouse spleens to perform liver functions, used gene editing to grow liver tissue without transplanting any cells, and even rebuilt thyroid tissue in animal models.Now, the next frontier is personal: using patient-specific induced pluripotent stem cellsto grow customized organs. “The spleen acts like a living bioreactor embedded in our bodies,” said Dong. “With minimally invasive B-ultrasound-guided delivery, we could one day cultivate custom-made organs on demand.”While the concept is promising, clinical use is still a few years away. However, after undergoing thorough safety testing, this work forces a re-evaluation of regenerative medicine and what we consider “non-essential.” The spleen, long overshadowed and often dismissed, might just be one of the body’s most underutilized resources — an internal bioreactor whose potential is only just starting to be realized.This article is not offering medical advice and should be used for informational purposes only.Article SourcesOur writers at Discovermagazine.com use peer-reviewed studies and high-quality sources for our articles, and our editors review for scientific accuracy and editorial standards. Review the sources used below for this article:Science Advances: Transforming the spleen into a liver-like organ in vivoScience Translational Medicine: Islet transplantation in immunomodulatory nanoparticle–remodeled spleensHaving worked as a biomedical research assistant in labs across three countries, Jenny excels at translating complex scientific concepts – ranging from medical breakthroughs and pharmacological discoveries to the latest in nutrition – into engaging, accessible content. Her interests extend to topics such as human evolution, psychology, and quirky animal stories. When she’s not immersed in a popular science book, you’ll find her catching waves or cruising around Vancouver Island on her longboard. #how #nonessential #spleen #could #regenerate

Researchers from Nanjing University of Aeronautics and Astronautics and multiple collaborating institutions are advancing the use of additive manufacturing (AM) to produce zinc-based biomaterials for biodegradable medical implants. Motivated by the need for temporary implants that naturally degrade in the body, thereby eliminating risks associated with long-term metal retention, the team investigated selective laser melting (SLM) and binder jetting as methods to process zinc and zinc oxide powders into patient-specific scaffolds for bone tissue regeneration. Their findings, published in Acta Biomaterialia, demonstrate the feasibility of fabricating porous zinc structures with tailored degradation rates and mechanical properties. The study addresses key challenges in fabricating zinc structures via AM, including the metal’s low boiling point, high reflectivity, and tendency to oxidize. These properties have historically complicated laser-based processing, limiting zinc’s use in load-bearing biomedical applications despite its attractive profile as a biodegradable, bioactive material. Zinc as a next-generation biodegradable metal for AM Zinc’s corrosion rate is slower than that of magnesium but significantly faster than iron, placing it in an ideal range for bioresorption over a clinically relevant time frame. It also exhibits inherent antibacterial properties and plays a role in osteogenesis. However, traditional manufacturing routes have struggled to produce complex, porous zinc scaffolds suitable for bone in-growth. Additive manufacturing enables the fabrication of patient-specific, lattice-based implants with fine control over pore geometry, strut thickness, and internal architecture. In this study, SLM was used to process zinc powder into porous structures, while inkjet printing of zinc oxide was followed by a post-processing step that included sintering and reduction to metallic zinc. Both methods demonstrated potential to overcome the design limitations of conventional manufacturing, with implications for orthopedic and craniofacial implant design.(i) Schematic of a typical Laser Powder Bed Fusion (LPBF) machine, illustrating the inert atmosphere within the construction chamber and the direction of gas movement indicated by blue arrows. (ii) (a) Typical Selective Laser Melting (SLM) process; (b) SLM process schematic showing the processing chamber and gas circulation system; (c) Parameters for processing. (iii) Setup for the Selective Laser Sintering (SLS) process. (iv) (a) Schematic of an Electron Beam Melting (EBM) machine. (v) Fused Deposition Modeling (FDM) process. (vi) (a) Diagram of laser powder Directed Energy Deposition (DED) systems; (b) Schematic of Wire Arc Additive Manufacturing (WAAM) equipment based on plasma arc welding. (vii) Schematics for the Binder Jetting (BJ) process. Image via Journal of Materials Research and Technology. Optimizing AM parameters for zinc processing SLM processing required fine-tuning to mitigate evaporation and reduce porosity caused by keyhole formation. The authors suggest that alloying zinc with elements such as magnesium, calcium, or silver may improve printability, mechanical performance, and degradation behavior. With optimized parameters, the team achieved scaffolds with compressive strengths in the range of cancellous bone and interconnected pores that facilitate vascularization and cell migration. Inkjet-based AM offered an alternative pathway, especially for producing lower-density structures with finer feature resolution. However, it introduced challenges related to shrinkage and sintering-induced defects. Despite these issues, both AM approaches enabled the fabrication of cytocompatible scaffolds that supported cell attachment and proliferation in vitro, meeting preliminary benchmarks for biocompatibility. In vitro antibacterial activity: Bacterial morphology on the surface of samples after co-culture with (a) S. aureus and (b) E. coli (arrows tips indicated dead bacteria with broken and incomplete bacterial cell walls); (c) Images of S. aureus and E. coli on TSA after co-cultured with samples; (d) Antibacterial rates calculated by colony counting method; (e) Antibacterial abilities of the samples after incubation with PBS for 3, 7 days. Image via Journal of Materials Research and Technology. Toward clinical translation and customized implants The paper positions AM zinc devices as candidates for temporary bone fixation, load-sharing scaffolds, and biodegradable stents. Unlike permanent metallic implants, these devices gradually degrade in the body, reducing long-term complication risks and eliminating the need for surgical removal. Additive manufacturing’s digital design flexibility further supports the integration of patient-specific anatomical data, potentially reducing recovery times and improving treatment outcomes. Looking ahead, the authors emphasize the need for further in vivo testing and alloy development to tune degradation rates and biofunctionality. Hybrid AM strategies, such as combining inkjet-printed sacrificial templates with SLM overlays, may allow for functionally graded materials and composite structures. Advancements in biodegradable implants Recent advancements in 3D printing of zinc-based biomaterials for biodegradable medical implants highlight the growing interest in utilizing AM to create patient-specific, bioresorbable metal implants. This trend is part of a broader movement in the field of AM of bioresorbable metals, where researchers are exploring materials like magnesium, iron, and zinc to develop implants that safely degrade within the body over time. One pertinent example is the work by engineers at Delft University of Technology, who have utilized extrusion-based 3D printing to fabricate biodegradable bone implants made of porous iron. Similar to zinc, porous iron is biodegradable and has potential as a temporary bone substitute that degrades as new bone regrows, thereby reducing the risk of long-term inflammation associated with permanent metal implants. The Delft team developed a purpose-built extrusion-based setup to overcome challenges related to the low biodegradation rate of bulk iron, achieving porous structures with enhanced biodegradability and mechanical properties suitable for bone healing. Another notable development is the research conducted by RWTH Aachen University, where scientists have been working on lattice structures manufactured from a zinc-magnesium (ZnMg) alloy using Laser Powder Bed Fusion (PBF-LB). These structures are designed to be patient-friendly and promote bone healing, with the ZnMg alloy offering a balance between mechanical strength and biodegradability. The researchers aim to develop bone-mimicking structures while gradually degrading in the body, eliminating the need for secondary surgeries to remove implants.As additive manufacturing continues to mature, zinc-based bioresorbable devices may offer a crucial link between materials science, digital fabrication, and personalized medicine.What 3D printing trends should you watch out for in 2025? How is the future of 3D printing shaping up? Subscribe to the 3D Printing Industry newsletter to keep up with the latest 3D printing news.You can also follow us on LinkedIn and subscribe to the 3D Printing Industry Youtube channel to access more exclusive content. Feature image shows a scanning electron microscope scan of a single laser scan cross-section of a tested nickel and zinc alloy structure. Image via Texas A&M University. Source: https://3dprintingindustry.com/news/additive-manufacturing-of-zinc-biomaterials-opens-new-possibilities-for-biodegradable-medical-implants-239427/?utm_source=rss&utm_medium=rss&utm_campaign=additive-manufacturing-of-zinc-biomaterials-opens-new-possibilities-for-biodegradable-medical-implants #additive #manufacturing #zinc #biomaterials #opens #new #possibilities #for #biodegradable #medical #implants

Apple’s Machine Learning team, in collaboration with researchers from Nanjing University and The Hong Kong University of Science and Technology, has announced an interesting 3D AI model called Matrix3D. This so-called Large Photogrammetry Model is able to reconstruct 3D objects and scenes from just a few 2D photos, but with a big difference from current pipelines. Here’s why this is a big deal. First things first: photogrammetry. It uses photographs to make measurements in order to create 3D models or maps. Currently, this process involves using different models for steps like pose estimation and depth prediction, which can lead to inefficiencies and errors. Matrix3D simplifies this by doing it all in one go. It takes in images, camera parameters (such as angle and focal length), and depth data, and processes them using a unified architecture. This not only simplifies the workflow but also improves accuracy. Even more interesting is how the model was trained. Researchers used a masked learning strategy, very similar to early Transformer-based AI systems that helped pave the way for the first versions of ChatGPT. They randomly hid parts of the input data during the training process, which forced Matrix3D to basically learn how to fill in the gaps. This technique is key because it enables Matrix3D to train effectively even with smaller or incomplete datasets. The results are impressive. With just three input images, Matrix3D can generate detailed 3D reconstructions of objects and even entire environments, which obviously could have very interesting applications for immersive headsets like the Apple Vision Pro. The researchers made the source code for Matrix3D available on GitHub, and published their paper on arXiv. They also created a website where you can watch more sample videos and even interact with a few point cloud recreations of objects and environments. Add 9to5Mac to your Google News feed.  FTC: We use income earning auto affiliate links. More.You’re reading 9to5Mac — experts who break news about Apple and its surrounding ecosystem, day after day. Be sure to check out our homepage for all the latest news, and follow 9to5Mac on Twitter, Facebook, and LinkedIn to stay in the loop. Don’t know where to start? Check out our exclusive stories, reviews, how-tos, and subscribe to our YouTube channel Source: https://9to5mac.com/2025/05/13/apple-study-3d-objects-from-images/ #new #apple #model #generates #scenes #from #just #three #images