#MCP / tool use

Researchers test AI agent clusters in Minecraft: Collaboration is prone to bureaucracy, higher win rate against GPT in battles

Researchers used 50 GPT-6.1 Sol agents to build a civilization community in Minecraft, but failed to develop due to bureaucratic tendencies; in a wake-up war test with 10 GPT-6.1 Sol against 10 Claude Opus 5.5, GPT agents adopted a more aggressive strategy, completing problem diagnosis and sharing solutions; in seven Mindustry battles, GPT agents won six and lost one, demonstrating the advantage of swarm mode.

Research indicates that collaborative design of System-1 and System-2 towards general artificial intelligence can reduce execution costs

Research proposes that by allowing the lightweight predictor System-1 to evaluate candidate interactions before execution, the execution cost of the resource-intensive executor System-2 can be reduced, ideally reducing the computational workload to 1/ρ_u times (where ρ_u is the utility-related interaction density), and this collaborative design method is applicable to scenarios such as training, inference, decision-making, and agent tool use.

Alibaba said that users of the China market version of Apple Intelligence can choose whether to enable the Qwen integration or not.

Alibaba announced that the China market version of Apple Intelligence will integrate with Qwen, allowing users to call Qwen's text and image understanding and content generation capabilities in Siri, writing tools, and visual intelligence, and they can choose whether to enable it or not. Related features will be open for testing with the latest iOS developer Beta version on November 4, and will be officially launched for Chinese users with the iOS 27.3 version in December. Alibaba stated that user requests and generated replies will not be used for model training.

WorkFactory: Transforming Professional Intent into Executable Tasks through Information Planning and Repair to Achieve Scalable Agent Training

WorkFactory builds coherent professional tasks through information planning and repair, producing the WorkPilot model, enhancing expert-level agent performance in professional work benchmarks. The framework starts from professional intent, constructs information support graphs to guide task instantiation, and identifies support failures and guides repairs through execution detection while preserving expected work.

Build generative UI applications with A2UI and CopilotKit agents

Google introduces its framework-agnostic open standard A2UI, allowing agents to describe interfaces declaratively in JSON, rendered by applications using existing components, and integrated with the AG-UI protocol of CopilotKit to connect to agent frameworks like LangChain, Google ADK, Mastra, Strands, etc.

Zhuji Dynamics releases FluxVLA Engine V2 technical report, one-click reuse of engineering capabilities makes embodied AI R&D more efficient

Zhuji Dynamics releases the FluxVLA Engine V2 technical report, achieving full-chain reuse of model training, evaluation, and deployment through unified configuration, modular decoupling, and system optimization. It supports 14 mainstream strategies, with inference path frequency improved by 2.31–9.64 times, validated in closed-loop execution on 6 real machines including ALOHA and dual-arm Franka. Online reinforcement learning boosts SmolVLA success rates by 10.0–15.4 percentage points.

EmbeddingGemma 2: A Unified and Efficient Cross-Modal Representation Model

EmbeddingGemma 2 is an open-source multimodal embedding model that unifies text, image, audio, and video representations in an efficient and resource-flexible design, and performs excellently in benchmark tests. It has established a new Pareto frontier in embedding benchmark tests for all modalities, outperforming professional models and other models that are two to twelve times larger in scale. The model supports selective modality encoder loading, adjustable visual token budgets, truncatable embedding dimensions, and quantized weights to adapt to resource-constrained environments.

Google announced the integration of third-party AI agents into Android Studio at the Android Dev Summit and released a new Gemini API

Google announced Android Studio Rabbit 2 at the Android Dev Summit, supporting more than 40 coding agents, including Claude, Codex, and Google Antigravity, to run within the IDE through the open-sourced Agent Client Protocol.

Meta Andromeda: Enabling Next-Generation Ad Retrieval through Model-System Co-Design

Meta 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.

Exowatt uses SageMaker serverless models to customize accuracy beyond state-of-the-art models

Exowatt 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.

Building enterprise-grade intelligent retrieval on Amazon Bedrock knowledge base using AWS CDK

Building 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.

The Omni-Woodpecker framework reduces cross-modal hallucinations in multimodal large models through evidence-prioritized rebalancing.

Omni-Woodpecker is a training-free framework that reduces cross-modal hallucinations in multimodal large models by identifying modal conflicts and rebalancing hidden representations, improving Qwen2.5-Omni-7B by 12.9 and 8.6 points on CMM and AVHBench, respectively, and consistently improving by 0.67-2.25 points on four general OLLM reasoning benchmarks, with the code open-sourced.

Xiaomi's open-sourced AI hardware HachimoDock debuts, with a DIY cost of about 150 yuan.

Factory Deployment Team: Pushing the boundaries of enterprise-level AI agent applications

The Factory Deployment Team delves into the real software environment of customers to solve the challenges of deploying autonomous agents, transforming engineering autonomy into corporate results through practices in companies like Blackstone and Nvidia. The team deploys agents on customer-owned infrastructure, achieving reasoning capabilities across large single codebases, and developing reusable product features such as code review and security review. The team is composed of highly autonomous engineering leaders responsible for strategic customer deployments and validating agent automation capabilities.

Nanbeige (Nanbeige) released the open-source intelligent agent programming model Nanbeige4.3-33B

Nanbeige (Nanbeige) released the 33B parameter dense large language model Nanbeige4.3-33B, supporting a context window of 524,288 tokens, leading in 17 out of 20 benchmark tests, including scoring 80.9 on Terminal-Bench 2.1, but lagging behind Qwen3.8-27B in SWE-Bench Pro.

iFlytek launched the Xingchen TaaS platform for the first time, making Tokens truly productive for thousands of industries.

iFlytek released the Xingchen TaaS platform, which transforms Token consumption into business value such as simultaneous interpretation translation and digital employees through intermediate capabilities like model scheduling and intelligent agent engines. It provides solutions for industries such as enterprise services, finance, and healthcare, ensuring that Tokens are not wasted. The Spark X2.6 edge model is open-sourced, with an expanded context window to 1 million Tokens and an edge-cloud routing accuracy of 96.1%.

Who should own the keys when writing code through an agent?

The amount and speed of code generated by the agent make manual review a bottleneck; the team needs to use tools as a substitute; the agent is only responsible for writing code without deployment permissions, deployment is controlled by pipelines; all data read by the agent is considered untrusted, the pipeline must ensure security through hash verification and short-term identity, and continuously monitor permission changes.

From KTIR to SpyreCode: Developing MLIR Dataflow Backend for IBM Spyre

Torch-spyre's Inductor determines how each kernel's work is allocated on Spyre's core and manages the residency of tensors between on-chip Scratchpad and off-chip memory, and the backend compiler translates these decisions into instructions that hardware can actually run.

MiHoYo, in collaboration with Tsinghua and Peking Universities, has released the open-source GameScaling game model, covering parameter scales from 0.8B to 27B.

MiHoYo, in collaboration with Tsinghua and Peking Universities, has released the open-source GameScaling game model, covering parameter scales from 0.8B to 27B. This series of 8 models is based on the Qwen3.5 visual language model for continued pre-training, with training data involving games such as Minecraft and Doom. The model generates subsequent instructions based on game screens and operation records, supporting real-time game interaction. There are two versions for each of the 4 parameter scales: S1 and S3, and they are licensed under Apache 2.0.

vLLM launches a push-based KV transmission to improve the TTFT and TPOT of distributed inference.

vLLM introduces a push-based NIXL connector, by moving post-pre-filled transmission coordination and setup out of the critical path, reducing time to the first token (TTFT) and time per output token (TPOT), enhancing application responsiveness.

Building an AI agent semantic layer on AWS using Strategy Mosaic and Amazon Bedrock AgentCore.

Strategy Mosaic, as a semantic layer, models C360 domains on Amazon Redshift and Amazon Aurora without ETL using two linked models. Its metric syntax captures the business definition of agent reasoning and exposes the model to AgentCore-hosted Strands agents via the Mosaic MCP server, supporting user identity propagation to achieve end-to-end row-level and object-level security.

Using X.509 PKI to protect AI agent identity on AWS

Keyfactor EJBCA integrated with AWS services provides identity authentication for AI agents through X.509 Public Key Infrastructure (PKI), addressing the issue of lacking built-in authentication standards when MCP connects to external tools.

Fine-tune GPT-OSS-20B on Amazon SageMaker HyperPod for multilingual inference

GPT-OSS-20B is an open-source inference model that thinks in English by default, and only 22% met the target language requirements in multilingual inference tests. By using supervised fine-tuning (SFT) and reinforcement learning (GRPO) on Amazon SageMaker HyperPod, the team can improve the model's inference capabilities in French, German, Spanish, and Italian, where SFT is responsible for primary behavior teaching and GRPO optimizes when there is room for improvement.

Goodfire researcher Nick Cammarata releases disruptive blog: The nature of neural networks might be easier to understand than previously thought

Goodfire researcher Nick Cammarata published a blog post stating that LLMs might be easier to understand than previously thought, causing shockwaves in the AI world. Cammarata believes that the mechanisms of neural networks may not be as difficult to decipher as previously thought, a view that is particularly noteworthy against the backdrop of frequent LLM proxy out-of-control incidents.

OpenAI has acquired Orchestra Bio, but hardware design remains a challenge for AI.

OpenAI has acquired Orchestra Bio. Current cutting-edge models like GPT-6 Astra, Fable 5.1, and Grok 4.6 can generate CAD, design PCBs, and describe assembly, but hardware design benchmarks like CAD Arena, EE Bench, and LuxoBench show that improvements are still needed. AI in hardware design faces the lossy problem of the interface between the digital and physical worlds, even if it can generate designs, a lot of work is still required.

DeepSeek V4.1 Flash 1017 released, US and China strengthen AI model regulation

MiniMax M3.1 Flash released: Next-generation model with improved performance and reduced cost

MiniMax M3.1 Flash is officially launched, with nearly a threefold increase in online inference speed and a 70% reduction in cost compared to the previous generation. The model performs excellently in benchmark tests for complex intelligent tasks (cowork), computer use, software development, and network security, ranking highly in the AutomationBench-AA for automated tasks, scoring 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2, 29.6 on ALE, and 82% on CyberGym.

vLLM optimizes Kimi K3: Fast and responsible model tuning practices

vLLM has built a unified CuTe DSL decoding path for Kimi K3 through intelligent agent-assisted development, achieving a batch 1 performance of 840 tokens/s on 16 NVIDIA GB300 GPUs, with a 1.5-2x increase in decoding throughput in the test configuration, and a 1.8x increase in AgentX workload P90 interactivity.