Hugging Face Daily Papers7575LaLa-World is a language-conditioned latent world model that achieves efficient model prediction planning and strategy initialization by combining simulation, robotics, and video data. The model adopts a language-conditioned interface of video generation models, enabling JEPA-style latent world models (LWMs) to learn across different domains and data sources.
The non-official PC port of 'Bloodborne' by Super Medo supports NVIDIA DLSS Frame Generation technology on the RTX 4070 graphics card, achieving stable performance above 220 FPS at 1080p resolution, which is over 7 times better than the original console version; with native 4K resolution enabled and DLAA, the average frame rate is about 40 FPS, and with DLSS 3 Frame Generation in quality mode, it exceeds 100 FPS, while in performance mode, it reaches about 130 FPS.
Google launches Agent Compute Engine (ACE) service, providing a secure, pauseable independent microVM environment for AI agents, supporting rapid wake-up (<500ms), initially used to support Gemini and other AI agents, and will be available for developers on Google Cloud in the future.
Hugging Face Daily Papers6565Policy-Mask learns a budget-aware diffusion step, network block, and channel selection strategy to extract compact and task-optimal features for discriminative visual tasks. Selecting intermediate features from a frozen diffusion model for discriminative tasks is complex because useful signals are distributed across denoising trajectories, network depth, and channels.
vLLM Blog5858vLLM provides zero-day support for MiniMax-M3.1-Flash on NVIDIA Blackwell, Rubin, and AMD MI350 series GPUs, performing excellently on agent workloads.
MiniMax open-source M3.1 Flash model weights, total parameters 428B, activation parameters 23B. Optimized through sparse attention, quantization-aware training, and Power-aware scheduling, inference efficiency is improved, processing over 100 trillion tokens in 72 hours after launch.
TechCrunch · AI7575MIT researchers placed 13 Eko agents running Claude Opus 4.7 on a virtual island for 30 hours without specific tasks, and they spontaneously created challenges and exhibited game-like behavior characteristics. After the experiment, these agents were generally superior to the control group that did not participate in the exploration. Eko learned to throw stones at a distance, stack structures, use environmental rules (such as the cliff rebirth mechanism) to achieve goals, and generated world knowledge unknown to the designers.
OpenAI announces full launch of the Decision API built on GPT-6 Luna, targeting questions with limited answers, assisting applications and agents in making near real-time decisions based on text, images, and structured inputs, with response speeds up to 10 times faster than calling GPT-6 Luna through the Responses API.
Why it matters: The original text provides the pricing, latency improvement, and single request limit changes for the Decision API, allowing readers to judge which type of agent workflow it suits.
Sakana AISakana AI has launched the Sakana Aquarium platform for building, executing, and running AI agents. It integrates organizational information through ontology, supporting the efficient construction and secure operation of AI agents. The platform includes three core mechanisms: ontology, governance, and agent execution base, achieving systematic organizational judgment criteria and task automation.
Hugging Face Daily Papers6565Hugging Face Daily Papers6565EvalResearchBench (ERB) is a benchmark for autonomous evaluation research, allowing AI agents to select or synthesize tasks, implement scorers, and revise within a fixed time and API budget, ultimately freezing an executable evaluator. 9 researchers and 13 candidate models were studied, and the evaluators generated scored approximately 75% consistency on 14 target benchmarks, below the 91% upper limit. Publicly designed tasks remain strong baselines, and the evaluator with the lowest execution cost achieved the highest pairwise consistency.
Modal Blog5858Sakana AI has released Namazu, a hybrid expert model with 1 trillion parameters, for Japanese business contexts, web search, and code execution. The Modal team achieved memory and throughput improvements for Namazu on the Modal platform through NVFP4 precision quantization, custom DFlash model training, and kernel-level optimization, and it supports real-time training of 3 billion tokens. After deployment, the model's inference speed increased by 40%, the acceptance rate improved by 43%, and the latency decreased by 52%.
Prime Intellect Blog3838Prime Mega MoE enables the DeepSeek Mega MoE kernel to be used for training by integrating backpropagation kernels, achieving up to 3.11 times faster MoE kernel speed compared to the default PyTorch implementation, and 1.34 times faster end-to-end training speed compared to the Prime-RL baseline.
Hugging Face Daily Papers2222RouteWeaver unifies heterogeneous LLM routing models through collaborative training and progressive exploration, improving accuracy in various reasoning tasks; its average accuracy reaches 61.4%, a relative improvement of 11.03% compared to the strongest baseline, and surpasses all comparison baselines in five domains.
Hugging Face Daily Papers6565T²Quant framework predicts accuracy sensitivity through minimal unlabeled calibration prompts, allocating activation precision within a fixed bit-width budget, improving existing quantization methods by 6.7 percentage points across 3 diffusion language models and 6 benchmark tests without increasing average accuracy budget.
Hugging Face Daily Papers6565RLSplat achieves linear reduction in latency, memory, and model size for feedforward 3D Gaussian splatting by assigning Gaussians through regional allocation and using local attention decoding. In zero-shot evaluation, RLSplat outperforms previous budget-controlled methods under all budgets, surpassing the strongest baseline even with a 4x reduction in Gaussian count, and achieves a 4.1x speedup and a 9.6x reduction in peak memory at low budgets.
Hugging Face Daily Papers4747Agent Mosaic quantified 11,240 responses from 3 LLM agents in simulated and open-source environments, finding that GPT-5.6-Sol had 7.08% of responses (95% CI) in simulations.
Hugging Face Daily Papers3838Heartbeat Thinking enhances the coverage and response speed of voice agents by continuing to prepare tool calls during voice updates. This technology allows voice agents to start preparing tool calls while the user is speaking, addressing the limitation of traditional methods where calls are interrupted by voice updates. In the FDBench v3 benchmark test, Heartbeat Thinking increased tool coverage by 16 percentage points and reduced average final response latency by 36.4%.
Microsoft's AI diagnostic system MAI-DxO achieved an accuracy rate of 83% in the New England Journal of Medicine's complex case simulation evaluation, which is 1.7 times the level of human doctors. The system simulates expert consultations by orchestrating multiple AI agents, with the standout model in the test being OpenAI GPT-5 and GPT-5.5.
Ollama releases4040Speech Ollama can use Phonon to generate speech on Apple Silicon, streaming audio via the /v1/audio/speech endpoint; the d3 decision model supports parameter sizes from 2B to 27B for classification, scoring, and other tasks; Flint expanded to Linux and Windows platforms, other improvements include error reporting and model capability transfer.
Liquid AI BlogSame story · 2 outletsLiquid RIPL returns complete generated responses in less than 50 milliseconds, supporting real-time requests on NVIDIA GPUs. LFM2.5-1.2B shows a 14.4–28.4 times lower p95 latency on RIPL compared to Qwen3.5-0.8B on vLLM, and Shopify has deployed it in production for search enhancement.
Modal BlogModal releases open-source Tinker compatible training server Spindle, supporting operation in Modal workspaces, replacing token billing with GPU time billing, customizable configurations, and supporting hundreds of concurrent training tasks. Spindle supports full-parameter training and multiple LoRA on various open-source models, with a distributed control plane and container auto-scaling at its core, suitable for teams needing custom training loops and cost reduction.
MIT Technology Review · AIEXL research finds that a company's AI competitive advantage depends more on the engineering systems around the model rather than the model itself; business load constraints should be defined first, then select appropriate models and service architectures, rather than doing the opposite; reliable, cost-effective, and large-scale AI application delivery needs to be achieved by measuring bottlenecks, targeted optimization, and building scalable architectures.
OpenShell 0.1.x released, providing a secure private runtime for autonomous AI agents, supporting a stable release pace, new isolation primitives, extended extension interfaces, and new APIs. Ensuring agents execute tasks without accessing real credentials, data, or networks through kernel-level policy enforcement (controlling file access, system calls, network connections) and formal verification (reviewing policy change risks).
LangChain Deep Agents releases4545`deepagents-code` 0.2.0 version now requires Python 3.14 and has removed low-value startup tests.
Amazon ScienceAmazon Science releases the open-source backend TorchNeuron, which translates PyTorch code into executable instructions for the AWS Trainium system, supporting existing PyTorch code to run on Trainium. It achieves efficient execution through optimizations like adaptive eager mode and multi-level cache, and uses an AI agent to screen 841 representative models and tens of thousands of tensor dimensions for benchmark testing.
TechCrunch · AI7070OpenAI has widely opened its Decision API to users for the first time, a model that reconstructs ChatGPT Luna to achieve faster inference and lower costs, reducing speed from second-level to millisecond-level.
Databricks Blog3838Fraud Analyst AI Coworker can detect telecom fraud in real-time on a single platform. The AI addresses the shortcomings of traditional fraud detection systems in terms of architecture through techniques such as cross-channel correlation, behavioral modeling, and device simulation. Its context window covers 6-7 system data sources, with a speed increase of 3.4 times, but the original text does not mention specific parameter scale, benchmark scores, price, and availability (open source/closed source/API).
OpenAI News3535OpenAI launches Decisions API, providing applications with fast, focused, and low-cost decision-making capabilities.
Hugging Face Daily Papers7070AgenticRoboRL framework improves robot planner performance via supervised fine-tuning and multi-round reinforcement learning, boosting the success rate of a 9B parameter planner from 18.1% to 71.0% on LIBERO-PRO without external memory, approaching Codex (GPT-5.5)'s 72.1% with external memory. The framework adapts to different execution mechanisms, linking recorded planner interactions with task outcomes to achieve a universal reinforcement learning training process through adapters.
AWS Machine Learning Blog2727IskayPet has built an enterprise knowledge assistant using Amazon Bedrock Managed Knowledge Bases, which can answer employee questions in four areas (retail operations, veterinary care, HR onboarding, and procurement), across three countries and four languages, without managing RAG infrastructure.
MarkTechPostSame story · 2 outletsLiquid AI has launched the RIPL real-time inference engine, serving the Liquid Foundation Model (LFM) in the request path, generating responses in less than 50 milliseconds end-to-end, and maintaining low latency under sustained peak traffic.
NVIDIA Technical Blog6565NVIDIA GPUDirect Storage can transfer data directly from storage devices to GPU memory. According to the 2025 Stanford AI Index, AI computing power doubles about every five months, and LLM training data size doubles about every eight months. GPUDirect Storage can improve data transfer efficiency.
AWS Machine Learning BlogAWS Machine Learning Blog5858The Amazon SageMaker Training Session API allows users to drive GPU-accelerated sessions from a local Python process, enabling algorithmic freedom through operations like forward_backward(), optim_step(), and sample(). SageMaker AI manages GPU resources.
AWS Machine Learning BlogSageMaker AI introduces model customization features without GPU management, allowing users to control post-training logic through training sessions without managing underlying infrastructure. Supports technologies such as SFT, DPO, RLVR, RLAIF, and multi-round RL, with options for LoRA or full fine-tuning.
AWS Machine Learning Blog3434Rescale utilizes Strands Agents and the AI-DLC framework to set up intelligent routing agents on AWS in five days and deploy them in production within weeks. The agent classifies end-to-end test failures at night by calling Claude Opus 4.8, retrieves historical failure patterns from the knowledge base, queries MCP logs for runtime evidence, and generates remediation reports containing file references, function names, and recommended fixes for engineers to review.
n8n BlogSelf-hosted LLMs allow you to control data storage location, suitable for sensitive workloads, but require bearing hardware, maintenance, and operational costs, while cloud APIs mostly handle these for you. A hybrid solution allows routing requests locally or in the cloud, without managing two separate systems.
