Pistis多模态大模型技术报告：IDRL后训练框架与PAH系统
原标题：Pistis Technical Report
AI 摘要
该技术报告介绍了Pistis多模态大语言模型系列，包含基于Qwen3.6的27B和基于Qwen3.5的9B两个规模，采用通用可扩展的后训练框架。框架先进行大规模多模态监督微调，再提出交错蒸馏与强化学习（IDRL）新范式，在单一训练循环中交替进行同策略蒸馏和强化学习，提升知识迁移、优化稳定性和长程智能体轨迹的信用分配。模型分为Pistis-Thinking和Pistis-Agentic两个变体，后者擅长多模态搜索，并引入系统级方法Pistis-Auto-Harnessing（PAH）自动改进推理编排。
正文节选
Pistis Technical Report Abstract We introduce the Pistis model family, comprising 27B- and 9B-parameter multimodal large language models built on Qwen3.6 and Qwen3.5, respectively, and developed through a general and scalable post-training framework. The framework first establishes a strong foundation through large-scale multimodal supervised fine-tuning (SFT). Building on this SFT foundation, we propose Interleaved Distillation and Reinforcement Learning (IDRL), a novel post-training paradigm t