Sign in to view Xiao’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Sign in to view Xiao’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Seattle, Washington, United States
Sign in to view Xiao’s full profile
Xiao can introduce you to 10+ people at Google DeepMind
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
2K followers
500+ connections
Sign in to view Xiao’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
View mutual connections with Xiao
Xiao can introduce you to 10+ people at Google DeepMind
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
View mutual connections with Xiao
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Sign in to view Xiao’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Activity
2K followers
-
-
Xiao Ma shared this🥳 We celebrated receiving Outstanding Paper from ICLR2025 last night! 🎉 Safety Alignment Should be Made More Than Just a Few Tokens Deep Xiangyu Qi, Ashwinee Panda, Kaifeng Lyu, Xiao Ma, Subhrajit Roy, Ahmad Beirami, Prateek Mittal, Peter Henderson Read the full paper here: https://lnkd.in/d85dDfV8
-
Xiao Ma shared thisIt was a pleasure to work on this paper. I like how we went to the very autoregressive nature of the LLMs and identified a very interesting root cause of many attacks. 🎉Xiao Ma shared thisExcited that our paper "safety alignment should be made more than just a few tokens deep" is recognized as an #ICLR2025 Outstanding Paper! We identified a common root cause to many safety vulnerabilities and pointed out some paths forward to address it! See Xiangyu's thread as he describes the paper and his thoughts on AI Safety in general!
-
Xiao Ma shared thisHere are 120 women to invite to your next AI panel. 🤗Xiao Ma shared this🚀 Since 2012, we have been showcasing brilliant women whose work in #VisualTech and #AI is reshaping business and society. We are thrilled to showcase 120+ women behind many of the technological innovations that we experience today: Lourdes Agapito, Zeynep Akata, Anima Anandkumar, Raquel Urtasun, Kavita Bala, Marian Bartlett, Regina Barzilay, Tamara Berg, Angjoo Kanazawa, Naila Murray, Ira Kemelmacher-Shlizerman, Octavia Camps, Duygu Ceylan, Rita Cucchiara, Dima Damen, Kristin Dana, Kate Darling, Tali Dekel, Ilke Demir, Deborah Estrin, Sanja Fidler, Chelsea Finn, Kristen Fortney, Carolina Galleguillos, Timnit Gebru, Molly Gibson, Michal Lipson, Georgia Gkioxari, Gaile Gordon, Deirdre Hanford, Kim Hazelwood, Judy Hoffman, Mary Lou Jepsen, Amba Kak, Fei-Fei Li, Devi Parikh, Anastasia Yendiki, Rosalind Picard, Joelle Pineau, Serena Yeung and many more! Click the link below to discover experts in machine vision, pattern recognition, and generative models, as well as leaders in #quantum hardware, #nanophotonics, and #semiconductors, which form the foundation for all visual tech and AI progress. Their contributions are extremely valuable to our ecosystem and the world! We want to thank our team, LDV Experts and our Community for nominating their colleagues, friends and peers. There are many more brilliant women – technologists, researchers, entrepreneurs – and we will keep updating this article as we meet more of them.120+ Women Spearheading Advances in Visual Tech and AI — LDV Capital120+ Women Spearheading Advances in Visual Tech and AI — LDV Capital
-
Xiao Ma shared thisHonored to join the board of this great nonprofit National Sawdust to promote diverse artists, and connect tech/AI with art and music.Xiao Ma shared thisNational Sawdust Welcomes 7 New Board Members On the eve of Its 10th Anniversary, the Brooklyn-based cultural institution grows in order to meet the evolving needs of 21st-century artists. National Sawdust has added seven new members to its Board of Directors: finance, technology, and advertising professional Kathie Choi; lawyer, corporate executive, and writer Chinwe Esimai; attorney Peter Faber; communications professional Alexandra Fenwick-Moore; performing arts producer Charles Letourneau; machine learning research engineer Xiao Ma; and philanthropy professional and community activist Teresa Toro. They will help lead the institution, which celebrates its 10th anniversary in 2024-25, into its second decade of impact as “an incubator for up-and-coming performers and composers” (WNYC) and “the city's most vital new-music hall” (New York Times). In this time of paradigm shifts in the performing arts, National Sawdust is growing to meet the moment. The Board expansion announced today follows a monumental accomplishment: National Sawdust's recent purchase of its $21 million venue in Williamsburg, Brooklyn. The acquisition gives the organization a more permanent place on the local, national, and international cultural landscape, and allows National Sawdust to direct even more resources toward today's most visionary artists, empowering them to experiment and make meaningful contributions to their field and to the world at large. Soon, the institution plans to announce a stellar 10th anniversary season that articulates its vision for the next decade—and beyond. National Sawdust Managing Director Ana De Archuleta said, “Empowering artists and contributing to New York City's vibrant performing arts ecosystem is an exciting opportunity and a joyful responsibility. As we embrace the complexities of today's world, we recognize the immense creativity and adaptability of artists and arts organizations. Together, we're navigating a landscape rich with new technologies and opportunities for connection and expression. Welcoming seven new board members, each bringing a treasure trove of knowledge and expertise, we are thrilled to enhance our support for artists, helping them thrive and make an indelible impact on our community.” https://lnkd.in/dHCT4BhZ
-
Xiao Ma shared thisHere is a clip of me talking about the need for a "hybrid" natural language and graphical user interfaces for GenAI at the @CODEConference at MIT. I was hinting at our #ExploreLLM work (https://lnkd.in/ed_fgRuF). The writeup should provide more context. https://lnkd.in/dPadAURB Thanks again David Holtz for moderating and the other panelist for the great discussion.Beyond ChatBots: ExploreLLM for Structured Thoughts and Personalized Model ResponsesBeyond ChatBots: ExploreLLM for Structured Thoughts and Personalized Model Responses
-
Xiao Ma shared thisThanks for having me!Xiao Ma shared thisVideos from the 2023 Conference on Digital Experimentation (hosted by the MIT Initiative on the Digital Economy) are now online! We've made available recordings of seven of our eight plenary talks (delivered by Hema Yoganarasimhan, Susan Murphy, Martin Tingley, Jean Pouget-Abadie, Jorge Guzman, Hannah Li, and Peng Ding), as well as our Practitioners Panel (with Tushar Shanker, Wenjing Zheng, Widad Machmouchi and James McQueen, moderated by Dean Eckles) and our Fireside Panel on Generative AI (with Daniel Rock, Brandon Stewart, Xiao Ma, and Payel Das, moderated by yours truly). There is some truly great stuff in here, so I hope that folks will take the time to watch some of this content, whether you missed the conference and are watching it for the first time, or were in attendance and want to reminisce about the old times. Thanks again to my co-organizers, Dean Eckles, John Horton, Sinan Aral and Alex 'Sandy' Pentland, as well as Carrie Reynolds, Aileen Ruël Menounos, David Verrill and everyone else at the MIT Initiative on the Digital Economy for their efforts putting together an amazing conference. Can't wait until CODE 2024!
-
Xiao Ma shared thisClosing the year with a bang! Same as many of you, I spent the year deep in LLMs. Here is my year wrapped, LLM-edition. 1. ♊ Contributed to the Gemini launch today![https://lnkd.in/eWtm9Ty5] 2. 🥂 Launched Bard (safely) earlier this year! 3. 🌶 Invented ExploreLLM, a new way of interacting with LLMs beyond chatbots with more structure and personalization. [https://lnkd.in/ed_fgRuF] 4. ❤️ Improved the diversity of LLMs. [https://lnkd.in/eAmcDP9s] 5. 🤔 Improved the moral reasoning ability of LLMs via thought experiments prompting. [https://lnkd.in/eqP6xbPK] 6. 🏅 Won a honorable mention for best paper at CHI. A better understanding of user trust in voice assistance and failure [https://lnkd.in/e5SXUxYZ]. 👩🍳 A few other exciting projects cooking and can't wait to share more in 2024. Have a great end of year everyone!
-
Xiao Ma shared thisXiao Ma shared thisThe language model reasoning literature (e.g., chain-of-thought) is predominantly focused on mathematical tasks with well-defined responses. What if we want to reason through more nuanced and complex topics with no clear right or wrong answer? ***Thought Experiments Prompting***
-
Xiao Ma liked thisXiao Ma liked thisI’m happy to share our latest perspective on the societal impact of AI in Nature. LLMs are increasingly being deployed across society and people interacting with them on a range of topics, including understanding healthcare, personal finance and so much more. As AI becomes more agentic, evaluations need to evolve accordingly. As part of our broader work on evaluating societal impact, our latest research explores how AI models can be geared to understanding the underlying reasons for why something is right or wrong, to ultimately make them more safe and helpful to people. This early work argues for an novel approach to ensure AI system don't merely mimic human behaviour, but instead respects different cultures and values around the world. We hope that this scientific framework will help move the industry toward more robust, scientifically grounded standards for AI development, and galvanise the AI research community to further explore ways to build AI that’s reliable and safe for everyone, everywhere. It was a privilege to collaborate on this with Julia Haas, Sophie Bridgers, Arianna Manzini, PhD (Oxon), and the rest of this brilliant team at Google DeepMind, as well as our academic partners. https://lnkd.in/es7waGc4
-
Xiao Ma liked thisXiao Ma liked thisIt's a smart model with big jumps on many capabilities. Proud that I had a chance to work closely as one of the primary post-training developers on this model in the last several months!Gemini 3.1 Pro: A smarter model for your most complex tasksGemini 3.1 Pro: A smarter model for your most complex tasks
-
Xiao Ma liked thisIt really is a very smart model, and getting here required our team to make some big "recipe changes" under the hood. I'm so proud of what we've built together and can't wait to see how you use it.Xiao Ma liked thisIt's a smart model with big jumps on many capabilities. Proud that I had a chance to work closely as one of the primary post-training developers on this model in the last several months!Gemini 3.1 Pro: A smarter model for your most complex tasksGemini 3.1 Pro: A smarter model for your most complex tasks
-
Xiao Ma liked thisXiao Ma liked thisAfter an incredible 14+ years at Meta, I’ll be moving on to my next adventure. It has been an honor to lead large engineering organizations across Instagram and Facebook during a time of massive growth. I’m deeply grateful to my teams and colleagues - together, we’ve tackled complex technical challenges and built products that serve billions. Leaving this community is bittersweet, but I’m ready for a new challenge. I’m thrilled to announce that I’ll be joining Google DeepMind to lead engineering for the Gemini app! We are at an inflection point in technology with AI, and I can’t wait to dive in with Chris Struhar, Jenny Blackburn, Josh Woodward, Demis Hassabis, and the team to help shape how these tools can help people everyday. I’ll be wrapping up at Meta over the next few weeks before starting in January. A huge thank you to everyone I’ve worked with over the years. Let's stay in touch!
-
Xiao Ma reacted on thisXiao Ma reacted on thisHello world, Gemini 3
Experience & Education
-
Google DeepMind
******** ********
-
******
******** ********
-
******* ****
******** ******** *********
-
******* **********
****** ** ********** ***** *********** ******* undefined
-
-
****** **********
******** ** ******* **** ********** *** *********** ***********
-
View Xiao’s full experience
See their title, tenure and more.
Welcome back
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
New to LinkedIn? Join now
or
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Licenses & Certifications
Languages
-
Chinese
Native or bilingual proficiency
-
English
Native or bilingual proficiency
-
French
Limited working proficiency
View Xiao’s full profile
-
See who you know in common
-
Get introduced
-
Contact Xiao directly
Other similar profiles
Explore more posts
-
Yizhe Zhang
Apple • 3K followers
We (w/ Shansan Gong, Ruixiang ZHANG, Huangjie Zheng, Jiatao Gu, Navdeep Jaitly, Lingpeng Kong) released a family of 7B diffusion language models, DiffuCoder, that specializes on code generation, with a focus on understanding and improving masked diffusion models. A core analysis of DiffuCoder is the autoregressiveness (AR-ness) score, a novel metric that quantifies the causal patterns in decoding, revealing how diffusion models break from strict left-to-right generation for more flexible, non-linear code planning. Recent advances in autoregressive (AR) models dominate code generation, but diffusion-based LLMs (dLLMs) like DiffuCoder offer a promising alternative, especially for complex programming tasks. DiffuCoder explores how these models decode differently—showing less global AR-ness in code tasks compared to math—and how temperature affects both token selection and generation order, unlike traditional AR models. We also introduce coupled-GRPO, a post-training RL method with a coupled-sampling scheme, to reduce performance drops during accelerated decoding, boosting parallelism and efficiency. We use a self-improvement pipeline that leverages AR-ness analysis, coupled-GRPO optimization, and evaluation on benchmarks like AceCode-89k to refine decoding strategies. This approach enables DiffuCoder to navigate diverse code generation pathways and enhance performance with modest computational overhead. Looking ahead, we aim to further leverage Reinforcement Learning to steer code generation through these decoding patterns, with the discrete nature of AR-ness scores providing a foundation for search-based strategies—ideal for the sparse rewards of optimizing complex code structures. Check out our full paper and code for a deeper dive! Paper: https://lnkd.in/gVWU3BDJ Code: https://lnkd.in/gmXTZ_6n Models: https://lnkd.in/gTcKCDr9 #MachineLearning #AI #CodeGeneration #DiffusionModels #NLP
220
5 Comments -
Sanghani Center for Artificial Intelligence and Data Analytics
2K followers
Longfeng Wu is an applied scientist intern at Amazon in Sunnyvale, CA, where she has been focusing on developing large-scale generative recommendation models based on semantic identifiers (IDs) aimed at fundamentally transforming how recommendations are generated and delivered. A Ph.D. student in computer science, Longfeng is advised by Dawei Zhou at the Sanghani Center for Artificial Intelligence and Data Analytics.
60
3 Comments -
Hadi Askari
University of California… • 1K followers
We’re excited to share our #NeurIPS2025 work: LayerIF: Estimating Layer Quality in Large Language Models Using Influence Functions! 📌 Our key idea: LLMs vary widely in how well different layers are trained. Instead of relying on heuristics or weight-based signals, LayerIF uses training-data influence scores to measure how each layer contributes to validation performance. This provides a data-centric, model-agnostic assessment of layer quality. 📈 We find: • Improved expert allocation for LoRA-MoE architectures • Data-driven layer sparsity allocation for pruning • A simple, generalizable pipeline for layer-wise diagnostics • Strong alignment between our influence-based signals and downstream performance If you’re working on LLM training, pruning, interpretability, or MoE systems, we’d love to chat. 📍 Come meet us at NeurIPS San Diego tomorrow at 4:30 PM during the poster session. 🔗 Paper link: https://lnkd.in/gVR284ws
61
2 Comments -
PyTorch
315K followers
Anyscale has contributed Ray to the PyTorch Foundation, joining PyTorch, vLLM, and DeepSpeed under open governance to advance scalable, distributed AI infrastructure. Janakiram MSV (Forbes) explains how this governance model reduces fragmentation and supports transparent collaboration across the AI ecosystem in this article: 🔗 https://lnkd.in/dwy4HpTC #PyTorchFoundation #Ray #OpenSourceAI #AIInfrastructure #DistributedAI
137
1 Comment -
AIRO TONY
BirdTech • 35 followers
A team from Tsinghua University has shattered the long-standing speed barrier in shortest path algorithms by designing a new solution that beats Dijkstra’s classical approach—without relying on sorting. Traditionally, Dijkstra’s algorithm finds shortest paths by repeatedly sorting nodes by distance, which imposes a fundamental speed bottleneck linked to sorting time. The new algorithm cleverly sidesteps this restriction by clustering boundary nodes and using selective exploration, reducing the number of candidates at each step and avoiding costly sorts. Their breakthrough also adapts Bellman-Ford techniques for directed graphs, combining randomized and deterministic innovations. This lets them solve shortest path problems faster than ever before—even for arbitrary graph weights—a feat considered impossible for 40 years. Their work won Best Paper at STOC and is set to reshape how developers approach pathfinding and graph optimization.Follow Coding Omega for future updates: #algorithm #programming #computer science
-
Dinesh Omapathi
I've been obsessed with… • 653 followers
I came across an incredible paper from Tsinghua University’s Duan Ran and team, which just won Best Paper at STOC 2025. They’ve cracked the code: a deterministic SSSP algorithm for directed graphs that beats Dijkstra’s classic O(m+nlogn) time by using a clever mix of bounded relaxations (like Bellman-Ford) and frontier clustering to avoid full sorting, achieving O(m⋅log2/3n). What does that mean for us in practical terms? 1. Scalability in action: Huge, sparse graphs—like large road networks, logistics simulations, or network topologies—could be processed much more quickly. 2. Simplified approach: No need for heavy priority queues. Less overhead, less complexity. 3. Better real-time performance: Great for environments where quick adjustments matter, such as navigation apps, real-time analytics, or distributed systems. 4. Updated thinking: Even a “solved” classical problem can evolve—proof that fresh thinking matters. The research is still early-stage and theoretical, but the implications for graph-heavy applications are exciting. Looking forward to seeing this inspire next-gen pathfinding tools and libraries. #SoftwareEngineering #Algorithms #GraphTheory #Innovation #STOC2025
6
-
Will Wolf
Gauntlet • 2K followers
Looking for a quick overview of Agentic RL for LLMs? Sharing a recent blog post with (interactive) flashcard-style summaries of techniques from "The Landscape of Agentic Reinforcement Learning for LLMs: A Survey (Zhang et al., 2026)," covering planning, tool-use, memory, self-improvement, and reasoning. Link in comments!
14
1 Comment -
Data Science Dojo
310K followers
📢 LLMs can reason. But can they act, adapt, and collaborate over time? Researchers from UIUC, with collaborators from Meta, Amazon, and Google DeepMind, just released a deep survey on Agentic Reasoning for Large Language Models — and it reframes how we should think about “reasoning” altogether. The core shift: This work argues that strong benchmarks ≠ real intelligence. Instead of one-shot reasoning, agentic reasoning treats LLMs as interactive agents that: - Plan over long horizons - Use tools and external systems - Learn from feedback and memory - Coordinate with other agents They organize agentic reasoning into three layers: - Foundational → planning, tool use, search - Self-evolving → reflection, memory, continual improvement - Collective → multi-agent collaboration and role coordination Why this matters right now: - If you’re building or working with agents, this explains why: - Prompt tricks alone don’t scale to real-world tasks - Memory and feedback loops are becoming first-class design primitives - Multi-agent systems are less about “more agents” and more about reasoning structure It also clearly separates: - In-context reasoning (orchestration at inference time) vs Post-training reasoning (RL and fine-tuning to internalize behaviors) That distinction alone changes how you design agent systems. Feels like a roadmap for where agentic AI is actually headed — not just what demos look cool on Twitter. #AgenticAI #LLMResearch #AIEngineering #Meta #Amazon #Google
37
-
Vasily Ilin
UW Math AI lab • 446 followers
Most existing #lean4 datasets contain only correct proofs. Models learn error correction with RL, that's expensive. With UW Math AI lab we release a dataset of 260k erroneous Lean proofs with - compiler feedback - reasoning trace - corrected proof Improvements in Error Correction: - Goedel 8B: 2x - Kimina 8B: 3x Paper: https://lnkd.in/gaybt4bd
117
2 Comments -
Darshan Jain
Sarva Suvidhaen Private… • 554 followers
A team from Tsinghua University has shattered the long-standing speed barrier in shortest path algorithms by designing a new solution that beats Dijkstra’s classical approach—without relying on sorting. Traditionally, Dijkstra’s algorithm finds shortest paths by repeatedly sorting nodes by distance, which imposes a fundamental speed bottleneck linked to sorting time. The new algorithm cleverly sidesteps this restriction by clustering boundary nodes and using selective exploration, reducing the number of candidates at each step and avoiding costly sorts. Their breakthrough also adapts Bellman-Ford techniques for directed graphs, combining randomized and deterministic innovations. This lets them solve shortest path problems faster than ever before—even for arbitrary graph weights—a feat considered impossible for 40 years. Their work won Best Paper at STOC and is set to reshape how developers approach pathfinding and graph optimisations. #algorithm #programming #computerscienc
4
-
Nina Peñaflor
LLM Arena • 1K followers
👉 My Key Takeaways from Chip Huyen's Recent Interview on Lenny's Podcast Chip Huyen is the author of the widely recognized "AI Engineering: Building Applications with Foundation Models". Link to the podcast: https://lnkd.in/dxZ-tFWX 💡 Importance of post-training. Pre-training gives you raw capabilities (next token prediction on massive data), but post-training is what makes the model actually usable. SFT on high-quality examples + RLHF. Fine-tuning should be your last resort, not first. Most problems can be solved with better prompts, better data, or RAG. 💡 Evals. You can't improve what you can't measure. Need multiple types: unit tests (does this specific prompt work?), integration tests (does the whole pipeline work?), regression tests (did we break something?), and user feedback loops. The hardest part isn't writing evals; it's maintaining them as your product evolves. 💡AI products. Reliability and UX matter more than models. Most AI product failures aren't about bad models: they're about reliability (API limits, latency spikes, poor monitoring) and UX (users don't understand how to use it, doesn't fit workflow). Building reliable platforms and talking to users constantly beats chasing SOTA models. Most insights come from watching users, not from benchmarks. 💡How to improve AI-powered apps. What people think improves apps: staying current on AI news, chasing newest agentic framework, obsessing over vector database choice, constantly evaluating model benchmarks, fine-tuning models. What actually improves apps: talking to users, building reliable platforms, preparing better data, optimizing end-to-end workflows, writing better prompts. Better prompt engineering beats switching models 90% of the time. A well-crafted system prompt, clear instructions, good examples (few-shot), and proper output formatting can transform a mediocre experience into a great one. 💡 Advice for builders. Start with user problem, not with cool AI technique. Use the simplest solution that works (often that's a good prompt, not a fine-tuned model). Build evals early. Focus on end-to-end experience. Don't fine-tune unless you've exhausted everything else. Don't treat AI as deterministic (it's not, you need to handle variability). Don't ignore data quality (garbage in, garbage out).
1
1 Comment -
Xinyuan Zhang
Meta • 1K followers
Releasing the Memory-QA Benchmark from EMNLP 2025 🚀 Excited to announce the official release of the benchmark dataset for our paper, "Memory-QA: Answering Recall Questions Based on Multimodal Memories"! I presented this work as an oral presentation at EMNLP 2025 in Suzhou last November. Since then, we’ve been working hard to make the resources available to the community. A huge shoutout to my co-first author, Hongda Jiang, for his incredible effort in getting this release over the finish line. 🔗 Code & Data: https://lnkd.in/eGrXX7Uz 📄 Paper: https://lnkd.in/enUXSbag ⚠️ A Note on Data Availability: For those diving into the repo, please note there are some small differences between the open-source release and the dataset used in the paper due to licensing restrictions: - Training Data: While we cannot release the specific QA pairs and memory entries used for training, we have provided all the training images. We also included detailed instructions and prompts in the paper so you can generate your own training data. - Test Sets: The test-l set is available (with a small portion of samples excluded), but the test-s set remains private. We hope this benchmark helps propel further research into multimodal memory QA and recall!
35
-
Charith Mendis
University of Illinois… • 1K followers
Excited to share that the ADAPT group (https://lnkd.in/gV4hZmS8) at UIUC will be presenting three papers at the upcoming OOPSLA 2025 conference next week. Here are some brief details of the papers. 𝐆𝐀𝐋𝐀: 𝐀 𝐇𝐢𝐠𝐡 𝐏𝐞𝐫𝐟𝐨𝐫𝐦𝐚𝐧𝐜𝐞 𝐆𝐫𝐚𝐩𝐡 𝐍𝐞𝐮𝐫𝐚𝐥 𝐍𝐞𝐭𝐰𝐨𝐫𝐤 𝐀𝐜𝐜𝐞𝐥𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐋𝐀𝐧𝐠𝐮𝐚𝐠𝐞 𝐚𝐧𝐝 𝐂𝐨𝐦𝐩𝐢𝐥𝐞𝐫: Most GNN acceleration systems either perform intra-operator optimizations (e.g., TACO, SparseTIR) or inter-operator optimizations (e.g., Graphiler), missing the synergistic benefits of performing both simultaneously. In this work, we propose a language and a compiler with two complementary IRs to effectively track and transform GNN code to perform both classes of optimizations (and more!) to achieve, on average, 2.55x geometric speedup across a wide range of GNNs, graphs, and systems. This work is led by the PhD student Damitha Sandeepa Lenadora 📄Paper: https://lnkd.in/gAPpm4P7 𝐒𝐏𝐋𝐀𝐓: 𝐀 𝐅𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤 𝐟𝐨𝐫 𝐎𝐩𝐭𝐢𝐦𝐢𝐬𝐞𝐝 𝐆𝐏𝐔 𝐂𝐨𝐝𝐞-𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐒𝐏𝐚𝐫𝐬𝐞 𝐫𝐞𝐠𝐮𝐋𝐚𝐫 𝐀𝐓𝐭𝐞𝐧𝐭𝐢𝐨𝐧: Transformer models are increasingly using sparsified attention mechanisms (e.g. Mistral) to handle large context lengths. In spite of their importance, we noticed that there does not exist general and high-performance code generation support for diverse sparse attention patterns used by different models. Aggravating this fact is that most sparse attention workloads exhibit moderate sparsity (20%-50% dense) and hence most sparse compilers and libraries are ill-suited to optimize these workloads. To address this gap, we introduce a new sparse format, ACSR, and an associated GPU code generation algorithm, SPLAT, that accelerates a diverse set of sparse attention workloads with a geometric speedup of over 2x. This work is led by PhD student Ahan Gupta 📄 Paper: https://lnkd.in/gzkdmuCW 𝐀𝐮𝐭𝐨𝐦𝐚𝐭𝐞𝐝 𝐕𝐞𝐫𝐢𝐟𝐢𝐜𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐒𝐨𝐮𝐧𝐝𝐧𝐞𝐬𝐬 𝐨𝐟 𝐃𝐍𝐍 𝐂𝐞𝐫𝐭𝐢𝐟𝐢𝐞𝐫𝐬: Abstract interpretation-based neural network verification techniques are popular in the literature. However, they are usually implemented in general tensor programming languages and their soundness is manually proven on paper. We for the first time introduce a new verification system that can automatically prove the soundness of verifiers specified using the NN certifier specification language we built earlier, known as ConstraintFlow. This allows rapid iteration of DNN certifier specifications with push-button feedback about the soundness. This work is led by PhD student Avaljot Singh, who is co-advised by Gagandeep Singh. 📄 Paper: https://lnkd.in/gpscbtKF Hoping to see you all at OOPSLA 2025 (https://lnkd.in/gH4j5ix7?) next week!
43
1 Comment -
Malvika Jadhav
University of Florida • 1K followers
Excited to share that our paper, "Uncovering the Deceptive Tactics of Stalkerware," has been accepted at ACM ASIACCS 2026! Stalkerware apps are designed to be invisible, and that invisibility turns out to be deeply engineered. Our work takes a large-scale longitudinal look at the Android stalkerware ecosystem, analyzing how these apps systematically under-disclose their surveillance capabilities, how a distinct class of apps within the ecosystem functions solely as installation guides for stalkerware, and how developer evasion strategies evolve in response to each new Android privacy update. A huge thank you to my co-authors Wenxuan Bao and Vincent Bindschaedler for making this work possible. I also want to thank the Coalition Against Stalkerware for their support in this work. Looking forward to presenting it and engaging with the community in Bangalore! #ASIACCS #MobileSecurity #AndroidSecurity #Privacy
39
-
Ankita Dhiman
NVIDIA • 6K followers
Just came across a breakthrough from researchers at Tsinghua University—they’ve developed a new shortest-path algorithm that can actually beat Dijkstra’s classic algorithm in certain cases. What’s new? Their method runs in O(m · log^(2/3) n) time, making it asymptotically faster than Dijkstra’s O(m + n log n) for graphs with non-negative weights. Why it matters: This could reshape fields like: Network routing Logistics & supply chain AI pathfinding Chip design Big data systems It’s a strong reminder that even “solved” problems in computer science can still spark surprising innovations. 📄 Read the full paper here: https://lnkd.in/gYwfcqCR Also I am preparing a vlog diving deeper into this algorithm— I’ll be posting that soon!
14
2 Comments -
IEEE Transactions on Circuits and Systems I: Regular Papers
5K followers
📄Check out our featured article, "A Generalize Hardware Debugging Approach for Large Language Models Semi-Synthetic, Datasets." ✍️ Authored by: Weimin Fu; Shijie Li; Yifang Zhao; Kaichen Yang; Xuan Zhang; Yier Jin; Xiaolong Guo We propose a directed, semi-synthetic data synthetic method that leverages version control information and journalistic event descriptions. To produce high-quality data, this approach utilizes version control data from hardware projects combined with the 5W1H (Who, What, When, Where, Why, How) journalistic principles. It facilitates the linear scaling of dataset volumes without depending on skilled labor. We have implemented this method on a collected dataset of open-source hardware designs and fine-tuned fifteen general-purpose LLMs to enable their capability in hardware debugging tasks, thereby validating the efficacy of our approach. 🔗 Continue reading on IEEE Xplore here👉 https://loom.ly/IJd7Nfs #TCASI #IEEE #IEEEXplore #CircuitsandSystems
-
Krithik Dhandapani
District of Columbia… • 654 followers
💡 Standing Out in Tech: Lessons from Lam Research’s Career Webinar by Justin Lai, PMP Lai and Audrey Choi Last week I attended a webinar where Audrey Choi, a former undergraduate student at UC Berkeley obtained a Master’s-level position at Lam Research. The webinar offered practical advice for students and early-career professionals interested in the semiconductor and tech industry. I came away with insights that I’m excited to apply as I continue developing my skills and projects. Some highlights: ⚡ Projects speak louder than degrees – hands-on experience can really make your resume stand out. ⚡ Be versatile – having multiple complementary skills helps you stand out from the crowd. ⚡ Network strategically – talking to professionals, seeking feedback, and making genuine connections can enhance both your resume and your career path. ⚡ Rejection is part of growth – keep applying, improve your materials, and show continuous development. Events like this are an excellent way to gain practical career advice, learn from experienced professionals, and connect with people shaping the industry. I’m looking forward to putting these insights into practice and continuing to grow in my career! #LamResearch #Semiconductors #CareerAdvice #Networking #ProfessionalGrowth #ProjectsMatter #SkillDevelopment #EarlyCareer
10
3 Comments -
Shashank Gaur
Lendbuzz • 910 followers
Exploring the Foundations of LLMs — Through Stanford CS336 Recently, I’ve been following Stanford’s CS336: Large Language Models course—an advanced graduate-level class now publicly available via Stanford Online. As someone deeply interested in machine learning systems and NLP, this has been an incredible opportunity to expand my understanding from the ground up. 📚 Course website: https://lnkd.in/gY_fzNU4 📺 Lectures (free on YouTube): StanfordOnline – CS336 🎯 Key Learnings So Far: 🔹 Tokenization isn’t just preprocessing—it's a critical design decision that affects model performance, vocabulary efficiency, and even generalization. Exploring approaches like byte pair encoding gave me a new appreciation for the intricacies behind input representation. 🔹 Profiling resource usage is essential. The lectures on GPU memory tracking and PyTorch profiling offered valuable insights into how large models interact with hardware—and how to reason about performance bottlenecks early in development. 🔹 LLMs are a full-stack problem. From tokenizer to transformer block to GPU kernel, this course emphasizes that building scalable models requires a systems perspective, not just a model-centric one. 🔹 Hands-on implementation matters. The code-oriented structure of the course pushes learners to go beyond theory and actually build core components of a language model—something I’m already integrating into my own learning process. 🙌 Grateful to Stanford University and the CS336 teaching team—for making this world-class curriculum openly accessible. It's a fantastic resource for any student or practitioner eager to understand the building blocks of today’s AI systems. I’m looking forward to diving deeper into topics like parallelism, custom GPU kernels, evaluation techniques, and alignment strategies in the weeks ahead. If you’re following the course too or working on similar topics, I’d love to connect and learn together! #StanfordCS336 #LLMs #MachineLearning #DeepLearning #NLP #Transformers #AI #PyTorch #OpenLearning #AIEducation #DataScience
6
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content