学歴
-
-
–
ゼミ・クラブ・サークル:Robotics
ボランティア経験
出版物
-
Google Scholar Profile
-
出版物を表示Please click the link to get a list of publications on Google Scholar.
類似するその他のプロフィール
その他の投稿を確認
-
Caiming Xiong
Meet AGUVIS: A pure vision-based framework for autonomous Computer/Mobile-use GUI agents, operating seamlessly across web, desktop, and mobile platforms without UI code. Key Features & Contributions - Pure Vision Framework: First fully autonomous pure vision GUI agent capable of performing tasks independently without relying on closed-source models - Cross-Platform Unification: Unified action space and plugin system that works consistently across different GUI environments - Comprehensive Dataset: Large-scale dataset of GUI agent trajectories with multimodal grounding and reasoning - Two-Stage Training: Novel training pipeline focusing on GUI grounding followed by planning and reasoning - Inner Monologue: Explicit planning and reasoning capabilities integrated into the model training - Generalization to OOD (Out-of-Distribution): AGUVIS demonstrates impressive generalization on OSWorld benchmark even without desktop trajectory data during training. Project Page: https://lnkd.in/guzkwpbB Paper: https://lnkd.in/gs7wXyJS GitHub: https://lnkd.in/gWeSkFmK Collaboratiion among Yiheng Xu Zekun Wang Junli Wang Dunjie Lu Tianbao Xie Amrita Saha Doyen Sahoo Tao Yu Caiming Xiong
6件のコメント -
Tae-Ho Kim
AI has significantly transformed the industrial landscape, yet many challenges remain unresolved. A common issue is that obtaining training data for specific tasks can be difficult, making it impossible for AI models to solve these tasks effectively. However, the capabilities of Vision-Language Models (VLMs) have traditionally been robust in zero-shot performance. We utilized a VLM to tackle one of the most challenging tasks: accident detection. Nota is not only applying VLM to the real world, but making the system efficient as well. We pursue the optimized Gen AI on the edge to solve diverse problems effectively. #genai #edgeai #vlm
-
Antoine SAUVAGE
I had some interesting time attending the Node-RED Con 2024 in Osaka, where several initiatives showcased how the Node-RED IoT framework could be effectively used across different industries. In a nutshell, I could mention: * The IoT and digital twin initiatives supported by the Shizuoka Prefectural Institute of Industrial Technology, for boosting the local businesses IoT adoption. * Seeed Studio's presentation about their "Hazard Response Mission Pack," an open-source AIoT toolkit. * An educational project utilizing Node-RED MCU (microcontroller units) to create a quick-response quiz machine for students. * Three-Up Technology work on IoT applications in manufacturing. * And more.. You can check out the full conference here in the attached video. Also, I have a confidence to say 🤫.. Some new AIoT product, seamlessly integrating your favorite IoT framework, is on its way => https://lnkd.in/gsj_Z8ae ! Looking forward to exchanging more with you very soon about our EdgeAI Station capabilities ! #noderedcon #noderedjp
-
Aleksa Gordić
Wake up - new scaling laws paper just dropped! "Scaling Laws for Precision" Multiple interesting findings: * Training larger models in lower precision can be compute optimal - i.e. for a given compute budget you want to train larger models in lower precision rather than smaller models in higher precision in order to achieve the optimal loss * Too much pretraining data can actually be harmful as you quantize such models to lower precision for inference. Llama models popularized the data saturation regime (training on many more tokens / param than recommended by Chinchilla scaling laws (20 tok/param)) as their goal was not to train a compute optimal model but instead to optimize for inference. These models are more suceptible to degradations due to aggressive quantization. * When number of params is held fixed, compute optimal precision increases approximately logarithmically with data, i.e. for larger models you probably don't want to train in low precision even if your GPUs/training stack supports it! We're hitting the quantization limits that have historically brought us a lot of speed improvements (going from fp32 -> bf16 -> fp8). Lots more details in the paper: https://lnkd.in/dC3_EVrs --- 🔔 If you're looking to hire best ML/tech talent check out my platform here: https://lnkd.in/duJZ3D9T post a job here: https://lnkd.in/df-Ukm8E
4件のコメント -
Patricia Friar
Check-out Kinnami's new blog "It's About the Data , not Just Devices - Resilient Data for Self-Driving Vehicles (and Other Autonomous Systems)". Self-driving cars make the headlines, however many autonomous systems at the computing edge face the same resiliency challenges. Kinnami AmiShare addresses the three pillars of data resiliency—data protection, security and availability. Our AI-driven policy engine provides centralized oversight for managing data resilience in complex distributed and autonomous systems, including self-driving cars, which need to operate at all times. #ResilientDataEverywhere https://lnkd.in/g_FtJdgv
-
Pierre Haren
Here is quite an interesting post on how AI LLMs can replicate themselves, given a few nudges: https://lnkd.in/g5ZCwapR And here is a direct pointer to the Chinese paper: https://lnkd.in/gB5BWGda We are getting really close to enabling bad actors to release an intelligent virus which could have very nefarious goals. If you want to be entertained, go to page 9 "Unexpected observations" to understand how Darwin also applies to LLMs. Overall, this is rather scary, and we are there because we let the LLMs learn from all the GitHub command shell codes. What have we done?
6件のコメント -
Ryan Rad
I'll be attending the IEEE Conference on Artificial Intelligence (CAI) on the sunny island of Singapore next week! If you're also planning to be there, I'd love to connect and discuss the latest AI advancements. 2024 IEEE Conference on Artificial Intelligence https://ieeecai.org/2024/ #AIforgood #responsibleAI
1件のコメント -
George Z. Lin
DIFFTransformer, from MSR, is a further improved architecture designed to enhance the attention mechanism in LLMs by improving focus on relevant context while reducing noise. Traditional Transformers often face challenges with attention allocation, which can lead to an overemphasis on irrelevant information and complicate the retrieval of key data. DIFFTransformer addresses this limitation through a differential attention mechanism that calculates attention scores by subtracting two distinct softmax attention maps, thereby minimizing noise and encouraging sparse attention patterns. Research indicates that DIFFTransformer surpasses conventional Transformers across various metrics, particularly in language modeling, long-context modeling, key information retrieval, and the mitigation of hallucinations. It achieves comparable performance with approximately 65% of the model size or training tokens typically required by traditional Transformers, making it an efficient option for applications that involve long sequences. In practical scenarios, DIFFTransformer improves accuracy in in-context learning tasks and shows enhanced robustness to order permutations, a common issue in traditional models. Additionally, it reduces activation outliers, which aids in better model quantization and more efficient implementations. The architecture maintains compatibility with existing Transformer structures and hyperparameters, ensuring straightforward implementation and training stability. The differential attention mechanism functions similarly to noise-canceling technologies, enhancing focus on critical information and reducing hallucinations in tasks such as question answering and text summarization. By allocating higher attention scores to relevant answers while minimizing noise, DIFFTransformer improves retrieval capabilities. DIFFTransformer signifies a meaningful advancement in LLM architecture, effectively managing context and diminishing noise, thus serving as a solid foundation for future developments in natural language processing. Arxiv: https://lnkd.in/eHD-SgbM
-
Hai Huang
Yet another innovative paper on State Space Models https://lnkd.in/eevz_GiE achieves arbitrarily low memory decay (meaning arbitrarily long memory) by filtering the input sequence with a precomputed (not learned) spectral filter. Recurrent models, such as SSMs, have a general form of x_t = A x_t-1 = A^2 x_t-2 = … = A^t x_0. Because of the power term A^t, they model nonlinearity well. However, this comes with a cost, ||A|| has to be < 1. Otherwise, the model is “explosive” and will grow out of control rather quickly. For any A satisfying ||A|| < 1, A^t goes to 0 very quickly, meaning the effect of x_0 on x_t diminishes rapidly, which is known as “memory decay”. This limitation is why SSMs may not outperform transformers in the needle-in-the-haystack test. That is until this paper. And the trick? View the signal in its waveform. If something occurs repeatedly without decay, it must be a wave. By using a spectral filter that transforms signals from the time domain to the frequency domain, the paper proves that Spectral SSMs can have a memory longer than any given L. And the best part? The spectral filter is precomputed, not trained. Like any other SSMs, a hybrid architecture leads to better performance. Intuitively, this is because some signals can be efficiently expressed in the frequency domain, while some other signals can be efficiently expressed in the time domain, and having both helps. #artificialintelligence #machinelearning #deeplearning
3件のコメント -
Ibrahim Sobh - PhD
✨ Exciting insights from Yann Dubois's recent lecture at Stanford University on building Large Language Models (#LLMs) 🟢 Components of LLMs: Focus on architecture, training algorithms, data collection, evaluation methods, and system optimization for efficient model deployment. 🟢 Pre-training vs. Post-training: Understand the classical language modeling paradigm and the emerging trend of fine-tuning models like ChatGPT to function as AI assistants. 🟢 Importance of Tokenization: Effective tokenization strategies are crucial for handling diverse languages and ensuring model performance. 💎 This lecture is a must-watch for anyone interested in #AI advancements! 👉 https://lnkd.in/dcevp3cJ
1件のコメント -
Dr. Fahim K Sufi
Coeus Institute's GERA, autonomously identified one of the most impactful News of the day under 'International Crisis' category. GERA's Innovative Rating algorithm RATED this event ("The epic bust-up between China and India could be ending") with a score of 1228.5. GERA's patented rating algorithm automatically identifies critical events, categorizes /classifies them, finds root-causes, discovers correlations, and predicts future events.
-
Antonio Montano 🪄
💥💥💥 FlashInfer: Efficient and Customizable Attention Engine for LLM Inference Serving Zihao Ye, Lequn Chen, Ruihang Lai, Wuwei Lin, Yineng Zhang, Stephanie Wang, Tianqi Chen, Baris Kasikci, Vinod Grover, Arvind Krishnamurthy, Luis Ceze Abstract Transformers, driven by attention mechanisms, form the foundation of large language models (LLMs). As these models scale up, efficient GPU attention kernels become essential for high-throughput and low-latency inference. Diverse LLM applications demand flexible and high-performance attention solutions. We present FlashInfer: a customizable and efficient attention engine for LLM serving. FlashInfer tackles KV-cache storage heterogeneity using block-sparse format and composable formats to optimize memory access and reduce redundancy. It also offers a customizable attention template, enabling adaptation to various settings through Just-In-Time (JIT) compilation. Additionally, FlashInfer's load-balanced scheduling algorithm adjusts to dynamism of user requests while maintaining compatibility with CUDAGraph which requires static configuration. FlashInfer have been integrated into leading LLM serving frameworks like SGLang, vLLM and MLC-Engine. Comprehensive kernel-level and end-to-end evaluations demonstrate FlashInfer's ability to significantly boost kernel performance across diverse inference scenarios: compared to state-of-the-art LLM serving solutions, FlashInfer achieve 29-69% inter-token-latency reduction compared to compiler backends for LLM serving benchmark, 28-30% latency reduction for long-context inference, and 13-17% speedup for LLM serving with parallel generation. 👉 https://lnkd.in/dERW4Ntm #machinelearning
-
Shubham Survase
Tesla Teases Robotaxi Ride-Hailing App Ahead of August Reveal 🌟 By Cointelegraph | FELIX NG Elon Musk provided glimpses into Tesla's highly anticipated robotaxi service during the Q1 earnings call, hinting at the forthcoming release in August. Despite this development, the company's stance on Bitcoin holdings remains unchanged. 🌐 🌐 🌐 👁️ Explore CodeTentacles Technologies in depth by clicking on this link: https://lnkd.in/dPVhrPGe 💬 Don't hesitate to connect if you see potential for collaboration. 😄 Found this post interesting? ✅ Feel free to drop a comment, share it, and hit the like button! 🔔 Make sure to enable notifications at the top of my profile to stay updated with my content. 🚀 Follow and connect with me for more updates! #Tesla #Robotaxi #RideHailing #AutonomousVehicles #ElonMusk #Bitcoin #Codetentaclestechnology https://lnkd.in/dZSF5vXp
-
Matt Rappaport
New Tesla Patent Reveals Clues to an Exciting Future of Humanoid Robots and AI Integration In a interesting discussion, Herbert Ong and Scott Walter dive deep into Tesla's recently published patent application WO2024072984A1 (link in comments). Entitled "Actuator and actuator design methodology," the patent primarily focuses on controlling robot movement through strategic actuator placement. However, the document also unveils important clues about the potential for end-to-end neural net training of humanoid robots, reminiscent of Tesla's approach to training its vehicles with FSD. Ong and Walter explore the concept of personalized humanoid skills. Imagine purchasing a bot with built-in capabilities, then having the option to augment and customize those skills through a "Skills Store" akin to Apple's App Store. This opens up a world of possibilities for tailoring robots to specific tasks and preferences. The discussion also touches on the potential for various robot form factors, each designed to excel at particular skills. Picture a tennis-playing robot that comes pre-trained with basic volleying instructions, but requires a "Skills Upgrade" to perform at a master level. This raises questions about the need to manufacture and train a diverse range of robot body types and sizes, mirroring how humans possess ideal physiological characteristics for different activities. Later, Ong and Walter discuss a scenario in which Tesla builds the hardware – including automobiles, humanoids, and other bot form factors – while xAI develops the brains. This arrangement opens up opportunities for licensing agreements between the two companies, that may help the companies navigate future antitrust concerns as these technologies are deployed at scale. The convergence of advanced robotics and artificial intelligence holds the potential to catalyze a profound societal transformation. As Tesla and xAI continue to push the boundaries of what's possible, I can envision a future where intelligent, adaptable robots become integral to various aspects of our lives – from manufacturing and transportation to healthcare, recreation, and beyond. The seamless integration of cutting-edge hardware and software will revolutionize the way we live and work, ushering in an era of unprecedented efficiency, productivity, and innovation. While there are still challenges to overcome and ethical considerations to address, the prospect of these combined technologies driving positive change on a global scale is awe-inspiring. #FrontierFrontier #Humanoids #AI #Robots
1件のコメント -
kaikai luo
1. 🚀 TextGrad uses text feedback for automatic differentiation, offering a new method for prompt optimization in large language models, effective in Q&A, molecular optimization, and radiotherapy planning. 2. 🔧 DSPy provides the algorithmic foundation for TextGrad to optimize LLMs' prompts and weights, reducing the need for manual adjustments. 3. 🌐 Automatic differentiation is crucial in deep learning, and TextGrad applies it to the text domain, optimizing through text feedback. 4. ✨ ProTeGi leverages natural language gradients and prompt editing to effectively optimize LLMs' prompts, finding the optimal configuration. 5. ❓ Despite TextGrad's innovation, its practical application may be limited by the performance of the teacher model and the need for extensive API calls. #AI #TextGrad #Innovation #DSPy #LLMOptimization #Algorithm #AutoDiff #DeepLearning #ProTeGi #PromptOptimization #AIOptimization #AIChallenges #TechLimitations
-
William W Collins
Efficient Quantum Gradient and Higher-order Derivative Estimation via Generalized Hadamard Test by Dantong Li, Dikshant Dulal, Mykhailo Ohorodnikov, Hanrui Wang, Yongshan Ding via Quantum sensing via zongquanzhou on Inoreader ([Global] Quantum Computing) URL: https://ift.tt/zTAh6GC arXiv:2408.05406v1 Announce Type: new Abstract: In the context of Noisy Intermediate-Scale Quantum (NISQ) computing, parameterized quantum circuits (PQCs) represent a promising paradigm for tackling challenges in quantum sensing, optimal control, optimization, and machine learning on near-term quantum hardware. Gradient-based methods are crucial for understanding the behavior of PQCs and have demonstrated substantial advantages in the convergence rates of Variational Quantum Algorithms (VQAs) compared to gradient-free methods. However, existing gradient estimation methods, such as Finite Difference, Parameter Shift Rule, Hadamard Test, and Direct Hadamard Test, often yield suboptimal gradient circuits for certain PQCs. To address these limitations, we introduce the Flexible Hadamard Test, which, when applied to first-order gradient estimation methods, can invert the roles of ansatz generators and observables. This inversion facilitates the use of measurement optimization techniques to efficiently compute PQC gradients. Additionally, to overcome the exponential cost of evaluating higher-order partial derivatives, we propose the $k$-fold Hadamard Test, which computes the $k^{th}$-order partial derivative using a single circuit. Furthermore, we introduce Quantum Automatic Differentiation (QAD), a unified gradient method that adaptively selects the best gradient estimation technique for individual parameters within a PQC. This represents the first implementation, to our knowledge, that departs from the conventional practice of uniformly applying a single method to all parameters. Through rigorous numerical experiments, we demonstrate the effectiveness of our proposed first-order gradient methods, showing up to an $O(N)$ factor improvement in circuit execution count for real PQC applications. Our research contributes to the acceleration of VQA computations, offering practical utility in the NISQ era of quantum computing.
-
George Z. Lin
Stability AI have introduced significant improvements through the development of Scaling Rectified Flow Transformers (RFTs) a key component of the StableDiffusion 3 architecture. This research enhances generative models for text-to-image synthesis by leveraging rectified flow models, which connect data and noise in a straight line, offering better theoretical properties, prompt stability, and conceptual simplicity compared to traditional diffusion models. The team shows improved noise sampling techniques for training rectified flow models, biased towards perceptually relevant scales, which demonstrate superior performance over established diffusion formulations. A novel diffusion-transformer based architecture is presented, using separate weights for image and text modalities, enabling a bidirectional flow of information. The backbone, MM-DiT, is designed to handle different domains using separate sets of trainable model weights, outperforming existing transformer-based diffusion backbones. The study also showcases the scalability of this architecture, with predictable scaling trends correlating lower validation loss with improved text-to-image synthesis. The largest models outperform current state-of-the-art models and improves upon text comprehension, typography, and overall image quality, as confirmed by human preference ratings. Arxiv: https://lnkd.in/e7dhHAt8
1件のコメント -
Antonio Montano 🪄
💥💥💥 LongRAG: A Dual-Perspective Retrieval-Augmented Generation Paradigm for Long-Context Question Answering Qingfei Zhao, Ruobing Wang, Yukuo Cen, Daren Zha, Shicheng Tan, Yuxiao Dong, Jie Tang Abstract Long-Context Question Answering (LCQA), a challenging task, aims to reason over long-context documents to yield accurate answers to questions. Existing long-context Large Language Models (LLMs) for LCQA often struggle with the "lost in the middle" issue. Retrieval-Augmented Generation (RAG) mitigates this issue by providing external factual evidence. However, its chunking strategy disrupts the global long-context information, and its low-quality retrieval in long contexts hinders LLMs from identifying effective factual details due to substantial noise. To this end, we propose LongRAG, a general, dual-perspective, and robust LLM-based RAG system paradigm for LCQA to enhance RAG's understanding of complex long-context knowledge (i.e., global information and factual details). We design LongRAG as a plug-and-play paradigm, facilitating adaptation to various domains and LLMs. Extensive experiments on three multi-hop datasets demonstrate that LongRAG significantly outperforms long-context LLMs (up by 6.94%), advanced RAG (up by 6.16%), and Vanilla RAG (up by 17.25%). Furthermore, we conduct quantitative ablation studies and multi-dimensional analyses, highlighting the effectiveness of the system's components and fine-tuning strategies. 👉 https://lnkd.in/d7PaajBQ #machinelearning
Aveek Dasという名前のその他のユーザー
-
Aveek Das
Bank of America | Fixed Income & Structured Products | NISM: Equity Derivatives, Research Analyst, Mutual Fund Advisor | MBA Finance - FORE’23
-
Aveek Das
IT Applications Developer@Kotak Mahindra Bank'23
-
Aveek Das
AWS | Azure | ITIL | PRINCE 2 | CSPO | An experienced business analyst with deep skills in business transformation through cloud native solutions. Prior Big 4 experience.
-
Aveek Das
LinkedInにはAveek Dasという名前のユーザーが46人います
Aveek Dasという名前のその他のユーザーを表示