**[Prime Intellect Rewrites Verifiers, Agent Training and Evaluation Now Modular Like Building Blocks]** According to monitoring by Beating, the AI training platform Prime Intellect has released verifiers 0.2.0, along with a preview of the next-generation Verifiers v1 architecture. Verifiers is an open-source framework for generating tasks, running, and scoring AI Agents, applicable for capability evaluation and reinforcement learning training. Prime Intellect has also open-sourced the model training framework prime-rl. Simply put, Verifiers defines tasks, tools, and scoring rules, while prime-rl trains models based on task results. Developers can download and deploy these two tools independently. Prime Intellect also operates the Environments Hub and Lab. The former is used for sharing and downloading ready-made training environments, while the latter provides hosted training services. Developers can either deploy the entire toolset themselves or directly use Prime Intellect's environment and computing power platform. The older version of Verifiers tied tasks and Agent execution methods together. In v1, these have been split into three components: - **Taskset**: Specifies what to do, what tools to provide, and how to score. - **Harness**: Determines how the Agent completes the task. - **Runtime**: Decides whether the task runs locally, in Docker, or in a remote sandbox. As a result, the same set of tasks can be executed by different Agents such as Codex, Kimi Code, or Terminus 2, and can run locally, in Docker, or in a remote sandbox. Developers no longer need to rewrite tasks and scoring rules every time they switch Agents or execution environments. Version 1 also supports recording branching processes such as sub-Agent calls and context compression, while saving Token IDs and log probabilities required for training. The new version is better suited for long tasks spanning hundreds of rounds and can directly use Agent execution trajectories for reinforcement learning. The upcoming 1.0.0 version plans to introduce multi-Agent environments and enhance support for environment frameworks such as OpenEnv, NeMo Gym, and OpenReward. [Original Link]