Core Insights - The article introduces Miles, a new reinforcement learning framework designed for enterprise-level large-scale MoE training and production workloads, developed by the LMSYS team as a fork of the lightweight framework slime [1][4]. Group 1: Framework Features - Miles inherits the lightweight and modular design principles of slime, making it a preferred tool for model scientists exploring algorithms [3]. - It implements Infrastructure-level True On-Policy to eliminate discrepancies between training and inference, achieving bit-wise consistency [5]. - The framework introduces speculative training through MTP Online Training, resulting in over 25% rollout acceleration [3][9]. Group 2: Memory Optimization - Miles incorporates advanced memory management techniques to maximize GPU performance without triggering out-of-memory (OOM) errors [8]. - It features online SFT for Draft Models, which enhances performance by preventing a decline in acceptance length during training [9]. - The framework includes mechanisms to avoid benign OOM errors and implements memory margin strategies to address NCCL-related OOM issues [10]. Group 3: Technical Upgrades - Miles supports full-stack optimization for SGLang and Megatron, ensuring compatibility with rapid iterations in training and inference frameworks [6]. - The modular design allows researchers to easily modify components like algorithms, data, sampling, and evaluation with minimal code changes [6]. - It provides a user-friendly interface for model scientists, allowing them to adjust important sampling or loss dynamics without delving into lower-level code [6]. Group 4: Future Development - The LMSYS team plans to enhance the FSDP backend for improved stability in large-scale distributed training [14]. - Future developments include independent rollout deployment, additional debugging tools, and formal mathematical verification for SFT/RL scripts [14]. - The roadmap also aims to support next-generation hardware like GB300 and expand capabilities for multi-modal training [18].

Core Viewpoint - The article discusses the interest of major venture capital firms, including Sequoia Capital and A16Z, in investing in the voice AI startup Sesame, which focuses on developing AI voice assistants and wearable devices [1][2]. Group 1: Investment and Valuation - Sesame is reportedly in discussions to raise at least $200 million, with potential valuation discussions reaching into the billions [2]. - A16Z has led Sesame's Series A funding round, although specific terms and timelines have not been disclosed [4]. Group 2: Technology and Product Development - Sesame has launched its voice assistants, Maya and Miles, which are accessible via smartphones and laptops, and aims to integrate voice assistants into glasses for hands-free communication [3]. - The voice AI technology is based on Meta's Llama large language model and has been enhanced through training on approximately 1 million hours of audio primarily in English [3]. Group 3: Market Context and Competition - The growing interest in AI that can communicate like humans is highlighted, with industry leaders like OpenAI and Meta also developing voice capabilities for their text-based AI products [2][6]. - Sesame may become an acquisition target for companies like Meta, OpenAI, Anthropic, or xAI, which are already working on integrating voice features into their AI systems [5].