PyTorch Blog5858PyTorch® 2.15 released, defaulting to the rewritten NCCL backend, supporting fault logging and rollback; torch.distributed adds NIXL backend for unilateral transmission; FSDP2 and pipeline parallel optimize memory communication control.
Meta EngineeringMeta Andromeda's Next-Generation Ad Retrieval Engine Achieves Near-Ranking-Level Model Performance at the Retrieval Layer, Processing Over 1 Billion Ads Per Card Per Second, System Throughput Reaches 100 Billion Times Per Second, 100 Times Higher Than the 2024 Version, and Brings a 2.4% Conversion Rate Increase for Advertisers.
Together AI Blog3838Together AI has launched an auto-repair feature for GPU clusters, which continuously monitors node health on both virtual machines and host machines. When a failure is detected, it recommends minimizing repair actions and provides evidence, allowing users to choose automatic execution or manual approval. This feature is now fully available, supporting passive health checks, repair recommendations, automatic repair policies (including fault and node-level authorization), idle wait control, upgrade tiers, and safety guards, and is performed via API and console; the preview version provides repair commitment support based on node tags for third-party detection systems.
Databricks Blog4545Databricks Vital Skills enhances Genie Code's knowledge in the healthcare and life sciences field through the open-source Agent Skills standard, achieving a score improvement in the BiomniBench benchmark test.
Google Developers Blog3434Google Released HappyDelivery, a Conversational Business Prototype Based on LangGraph and Gemini, Implemented on GCP; Ingress Layer Processes WhatsApp Inputs via GKE/Cloud Run and Cloud Pub/Sub
Databricks BlogDatabricks improves document extraction accuracy by building a document extraction process that includes a review step, using ai_extract combined with JSON schema and global instructions, and leveraging MLflow Prompt Registry for versioned configuration. Finally, it validates the improvement effect on a fixed set of manually labeled data, enhancing the overall performance of the system.
Google Research5858Google Research proposes a sleep framework that enables LLMs to achieve continuous learning through alternating active (awake) and sleep phases; the sleep phase utilizes the MoE and Hope architectures for parameter expansion and knowledge distillation, integrating short-term memory into long-term parameters, supporting the retention of long-context knowledge up to 10M tokens and enhancing complex reasoning capabilities.
AWS Machine Learning Blog3434Exowatt customizes task-specific fine-tuned models Qwen3-1.7B and Qwen3-14B for edge and cloud deployment targets through Amazon SageMaker AI serverless models, achieving accuracy surpassing state-of-the-art models while reducing costs up to 49 times and shortening the model development cycle from months to days.
AWS Machine Learning Blog3434Property Finder discovered that 96.8% of the labels in its evaluation dataset for visual AI workloads migrated to Amazon Bedrock were predictions from a single model rather than human validation, hindering objective comparison of new models. The company built a trustworthy benchmark dataset by independently labeling 356 images with human experts and, after 27 iterative optimizations with evolutionary prompt tuning, reduced costs by 86% while maintaining an 86% accuracy rate.
AWS Machine Learning Blog3434AWS SageMaker AI introduces SOCI-indexed images functionality, removing container image downloads from the critical path, reducing SageMaker AI training and processing task startup times from minutes to seconds. This feature is achieved by pre-indexing AWS Deep Learning Container (DLC) or adding indexing to custom images, requiring no changes to workload definitions. Benchmark tests show it can reduce GPU idle time by 90% and improve efficiency 8-46 times.
NVIDIA Newsroom5858NVIDIA Vera Rubin NVL72 and NVIDIA Vera CPU are now officially available on Oracle Cloud Infrastructure (OCI), with plans to introduce HGX Rubin NVL8 systems to help organizations accelerate cutting-edge model training, improve inference efficiency, and expand agent workloads.
Hugging Face BlogHugging Face releases Old Ideas, Novel Problems study, evaluating the reliability of six LLMs (GPT-5.1 to 5.4, Claude Sonnet 4.5 to 4.6) in judging the reliability of research innovation.
NVIDIA BlogC-BRAIN and NVIDIA are using AI to change the pace of dementia research. They are accelerating the development of new therapies by integrating previously unpublished Alzheimer's research data, including unpublished results and controlled proprietary data. C-BRAIN employs a hybrid centralized and federated architecture to protect data privacy, deploying 64 NVIDIA data center GPUs and cloud nodes, and plans to introduce the NVIDIA Blackwell system to support its expansion goals.
Ant Group inclusionAI · Hugging Face new models5858inclusionAI/Ling-3.1-flash-1019 released, 560B parameter size, 25B active parameters/Token, 1M context window, improved reasoning ability, AA-Briefcase score 1524 Elo, AutomationBench-AA 68%, Artificial Analysis Intelligence Index v4.3.2 44 points. Supports SGLang and vLLM inference.
Ai2 BlogAi2 has released a study titled 'Old Ideas, Novel Problems,' evaluating the reliability of six LLMs (GPT-5.1 to 5.4, Claude Sonnet 4.5 to 4.6) in judging the innovativeness of scientific research. The study found that fine-tuning prompts can lead to accuracy differences of over 50%. The research constructed a benchmark dataset containing 154 human-generated and 154 AI-generated scientific ideas, confirming that models have insufficient stability in judgment and require more reliable methods to evaluate new ideas proposed by AI.
Ant Group inclusionAI · Hugging Face new models5858SingProbe is a streaming safety probe based on Qwen/Qwen3.8-Flash-Next, which scores query intent, response unsafety, and hallucination risk at each step of generation by reusing the hidden states of the base model, with less than 0.5% overhead during decoding.
Ant Group inclusionAI · Hugging Face new models5858inclusionAI releases a lightweight streaming security probe SingProbe based on GLM-5.3-Flash, which scores query intent, response unsafety, and hallucination risk in real-time by reusing the hidden states of the base model.
Ant Group inclusionAI · Hugging Face new models3838inclusionAI/Ling-3.0-flash-Sante-fp8 released FP8 quantized version, retaining 124B parameters (5.1B/Token) and BF16 performance, reducing memory and improving deployment efficiency, supporting medical inference, research, and tool-assisted workflows, requiring professional manual review. Stable performance on benchmarks like Healthbench, ReDisQA, achieving efficient inference through block FP8 quantization (128x128 blocks).
Databricks BlogLakeflow Connect supports wikis, project management tools, and AI base models through native connectors, enabling downstream AI agents to obtain complete business context. The platform provides fully managed native connectors that support SaaS applications, databases, and file sources, with data directly imported into the Databricks platform, and incremental import of Unity Catalog tables through point-and-click UI or API configuration.
LangChain Deep Agents releasesThe deepagents-code==0.2.2 version of LangChain Deep Agents has been released, including security and performance improvements. New features include GPT-6.1 Terra support, Chinese text input to chat, plugin update detection, etc.; bug fixes include project .env file trust value settings, terminal UI link limits, Git option limits, etc.; performance has been improved by reusing parent process context, optimizing server startup processes, and streamlining transcription rendering, etc. It is recommended to upgrade as soon as possible.
Databricks Blog2222Databricks Free Edition allows teachers and students to run full data or AI courses in the browser for free, without a cloud contract or license management, suitable for students and educators worldwide. The platform offers features such as collaborative notebooks, Spark, Delta Lake, Lakeflow, Databricks SQL, and Genie code assistant, supporting courses from introductory programming to AI projects. Skills are applied directly to industry-standard tools, with no conversion costs between production environments and teaching sandboxes.
Apple Machine Learning ResearchThe GHARP method animates 3D heads in real-time using a small amount of input images and expression signals, divided into an offline identity phase (constructing geometry and appearance representations) and a runtime animation phase (predicting expression residuals). This method performs animation in the semantic structure latent space of the pre-trained reconstruction model, limiting the expression change space to achieve efficient residual prediction, and achieves SOTA quality on the Ava-256 benchmark, with a 13x speedup on A100 GPUs and an 8x reduction in the number of Gaussians.
PyTorch Blog3434Nemotron 3.5 achieved the highest streaming speech recognition accuracy on Google Pixel 11 using ExecuTorch and SME2, reaching up to 11.7% WER (with 1,120 ms audio blocks and QC8W precision), which is 3.5x faster than FP32 but with increased latency; shorter audio blocks (e.g., 80 ms) can reduce latency but accuracy drops to 15.2% WER.
AWS Machine Learning Blog3838Amazon SageMaker SDK v3 released, training anomaly detectors and defect classifiers for each camera through dual-branch SageMaker Pipelines, and setting quality gates to prevent underperforming models.
Claude · YouTube5050Anthropic engineer Doug shares how to use Claude Docs in stand-ups and release reviews. Claude collaborates in documents, automatically extracting data, validating arguments, and responding to questions in comments. The tool is designed for team collaboration, supporting real-time analysis within the document.
AWS Machine Learning Blog4848Building enterprise-grade intelligent retrieval on Amazon Bedrock knowledge base using AWS CDK, providing multi-round planning and tool-driven retrieval through the reasoning agent and AgenticRetrieveStream API, returning referenced answers. The solution deployment includes a CDK application with Amazon S3 corpus, knowledge base, and reasoning agent, supporting semantic routing and multiple retrievals, and built-in Amazon CloudWatch monitoring and evaluation features.
IBM Research Blog5858IBM and the PyTorch Foundation announce that Spyre accelerator will become the first-class backend for PyTorch and support multi-accelerator scheduling. The torch-spyre compiler, Spyre vLLM plugin, and continuous integration efforts enable models and workflows to run on Spyre, an architecture optimized for inference, suitable for IBM Z, LinuxONE, and Power systems.
Ant Group inclusionAI · Hugging Face new models5858inclusionAI/Ling-3.1-flash-base is released, a sparse MoE architecture large model with 560B total parameters and 25B active parameters. It adopts a hybrid linear attention mechanism (63 KDA + 9 Gated MLA) and performs excellently in 47 cross-domain benchmark tests spanning eight fields. It supports continuous pre-training, domain adaptation, and post-training research. The model is trained with FP8 precision and released with BF16 precision, following the MIT license.
vLLM BlogFeatured6060The vLLM team used P/D separation and Wide-EP to serve DeepSeek-V4.1-Flash on a single NVIDIA GB300 NVL72 rack, achieving 954K tokens/s per GPU and 196 tokens/s for single-user P90 on SemiAnalysis AgentX, which is 7.9 times the early single-node results.
Why it matters: The vLLM team has released per-item optimization and measured data for serving DeepSeek-V4.1-Flash on GB300 NVL72, which can be compared with DeepSeek's own engine DeepInfer.
vLLM BlogFeatured6262vLLM team releases the second part of GLM-5.3 optimization, deploying on 16 H200s with 1 pre-filled node and 1 decoding node, using PCP to split pre-fill by token and DCP to split KV cache by token, allowing 1M context to fit into the pre-fill node
Why it matters: The vLLM team has dissected how PCP, DCP, and Hybrid HiSparse can be combined on 16 H200s, allowing readers to understand the memory and pre-fetch bottlenecks of long-context deployment.
Databricks Blog5858Spark's declarative pipelines are 34 times faster than imperative pipelines in 22 data engineering tasks, halve the code volume, and have fewer errors. Declarative pipelines capture structural errors in advance through pre-execution validation and offload repetitive tasks like sorting and result saving to Spark, while imperative pipelines need to run on the cluster until the error step to detect them. The two methods perform equally in terms of logical error rates, but declarative pipelines have more concise and consistent code, making them easier to maintain.
AWS Machine Learning Blog5858Amazon SageMaker AI trains models to perform sequential tasks using Multi-Turn Reinforcement Learning (MTRL) without requiring correct trajectories, only defining result scoring. Taking the Wordle game as an example, it increases the solving rate of the GPT-OSS-20B model from 61% to 84% (default inference strength), with a cost of about 339 USD.
LangChain Deep Agents releases3535deepagents-code 0.2.1 released, default Anthropic Auto classifier set to Claude Sonnet 5.5, new skill range MCP tool support added.
Claude · YouTubeFactory uses Claude's cutting-edge models to build AI coding agents Droids to achieve autonomy in enterprise-level software engineering. Claude handles long-term tasks with large code bases, enabling teams to trust agents with less supervision. Building a fully autonomous software 'dark factory' still requires overcoming many obstacles.
Hugging Face BlogLitPivot helps researchers adjust their research direction by reading previous literature, dividing research ideas into problems, contributions, and evaluation plans, and providing literature-based criticism and revision suggestions for each section. In a laboratory study involving 17 researchers, domain experts rated the literature foundation of novel ideas developed using LitPivot better than ideas developed using text editors and AI chat tools. Participants also reported stronger literature comprehension and improved understanding of the novelty and practicality of their ideas.
Hugging Face BlogHugging Face releases torch-neuronx model supporting query Space, detecting AWS Trainium compatibility. This Space supports querying whether mainstream models (such as Qwen3-0.6B) with a scale of 175B parameters can run on Trainium without modification, and provides running scripts.
Tencent Hunyuan · Hugging Face new modelsTencent Mixuan released the WeVisOne-VL-4B model, built on Qwen3-VL, integrating a 9.6M parameter pixel decoder, supporting object-level perception (open-vocabulary object detection, pointing localization, counting, pointing and OCR) and pixel-level perception (pointing segmentation and promptable alpha friction), suitable for research tasks such as visual localization, pointing, counting, OCR localization, pointing segmentation, and image friction.
