ReOPD 연구·Grok Imagine·Alibaba 토큰플랜과 희귀 도서 스캔 논쟁

Very cool paper from Microsoft. The idea is to train agents on replayed teacher trajectories instead of live environment rollouts. On-policy distillation for agentic tasks is expensive because every update needs fresh student rollouts through the environment plus teacher queries at each visited history. New research from Microsoft Research and the University of Amsterdam introduces ReOPD, which reuses pre-collected teacher trajectories as replayed prefixes. The student acts at selected steps while the teacher supplies dense per-step supervision without executing anything new. The paper also names a real pathology in multi-turn distillation, the prefix trap. Pushing histories toward the student's own distribution makes them more relevant to the student and simultaneously drags the teacher onto states where its targets are unreliable. Two-sided distribution shift between student occupancy and teacher reliability. ReOPD treats this as reliability-aware prefix distribution design and implements it with a step-decaying sampling schedule that emphasizes early, lower-shift prefixes. Across math reasoning with Python and search environments, over multiple teacher and student scales, it preserves or improves accuracy, uses zero tool calls during student training, and runs at least 4x faster per rollout. Paper: https:// arxiv.org/abs/2607.04763 Learn to build effective AI agents in our academy: https:// academy.dair.ai

Why manage separate AI tools for the script, the image, and the video? The Alibaba Cloud Token Plan gives you one shared credit pool across supported models and tools, with visibility into usage and access to newer models like Qwen3.8-Max-Preview, HappyHorse1.1, DeepSeek V4, and GLM-5.2. One plan for every modality, to build more, spend less. Get started from just $4 in your first month. Explore the Token Plans: https:// click.alibabacloud.com/m/20000000854/ #AlibabaCloud #TokenPlan #Qwen #Wan #HappyHorse #GenerativeAI #AIContent #MultimodalAI

Elon Musk

@elonmusk

Grok Imagine

yeah this kind of sucks but also it's not like you were ever going to read that rare book anyway. at least now it dies an honorable death in the service of creating Superintelligence, and the commies will make sure to distill and distribute its benefits to all!

FilthyTrading

@Filthy_Trading

model optionality and harness control are mostly a refactor. the learning loop is the actual work, and i think most teams claiming to own their intelligence have the first two and not the third

Susmita Manna

@SusmitaCodes007

Read this... You will understand which part of AI should a company own rather than owning whole and building it from scratch.