HF RL Explorer

AgentTrace is the bridge between framework-specific spans and the conversation history that callers feed…

AgentTrace is the bridge between framework-specific spans and the conversation history that callers feed…: a task in MiMo-V2.6-RL-oss: Agentic RL Environments (MiMo RL release). Its current spans to messages() projection is lossy: it removes every message whose role/content appeared earlier, so a…

The task

Its current spans_to_messages() projection is lossy: it removes every message whose role/content appeared earlier, so a user repeating the same words disappears; cumulative LLM snapshots can also be replayed in the wrong shape; and tool spans are rendered as assistant prose…

Part of XiaomiMiMo/MiMo-V2.6-RL-oss.