Tag
Laminar (YC S24) 发布了 flow-1，一个通过强化学习训练、专门用于检测智能体（agent）轨迹中静默故障的模型，在检测质量上匹配 GPT-6-sol 但成本低 23 倍，并集成在 Signals 分析框架和其开源可观测性平台中。
MemScope 是一个由中国科学院提出的框架，通过工具使用的 Patch Agent 检查原始执行轨迹、运行时内存状态和现有架构代码，实现对智能体记忆架构的证据驱动、局部化原位进化，在 GAIA、xBench 和 ALFWorld 上均取得显著提升。
Arize Phoenix announces updates including faster trace analysis with natural-language filters and an expanded REST API for managing retention assignments and model providers.
Phoenix has a built-in Pixie assistant that helps users quickly filter out silent failure traces where agent spans have errors but model responses are normal, greatly improving trace reading efficiency.
HALO is an open-source desktop app that uses reinforcement learning from model-based (RLM) techniques to debug and optimize AI agent traces locally, providing analysis and actionable recommendations.
Braintrust's Topics feature uses LLM summarization to make production agent traces tractable for clustering and classification at scale, inspired by Anthropic's Clio approach.
This paper investigates a critical disconnect in trace-based knowledge distillation for LLMs, revealing that semantically correct Chain-of-Thought traces are not reliably correlated with correct final answers and that traces optimized for model performance are often least interpretable to end users.