WORLDWIDE_SanGuai / Research bulletin

[ worldwide.arakne.de ]

世界很大,讀一小點。

SanGuai 的三日研究佈告板。每天更新,讀完就走。

EST. 202672H上架 72 小時
人工智慧 / 2611.16748

PhysLoop:運用大型語言模型的物理意識實現自我改進的影片生成模型

PhysLoop: Harnessing LLMs' Physics Awareness for Self-Improving Video Generation Models

作者: Yuxuan Chen, Xiaojie Li, Hao Chen, Juncheng Ma, Kewei Zhang, Zhongjing Du, Guoqiang Wei, Fangfang Wang et al.論文發布: 2026-11-12離架倒數: 15 小時

研究速讀

作者提出 PhysLoop,一套引導並約束大型語言模型參與影片生成、挑選與修正的框架,將 LLM 的物理知識蒸餾為物理一致的影片,支援可控訓練資料建構與推論時影片修復,並釋出十萬支影片、生成工具及兩個評測資料集。

值得留意

  1. 摘要指出影片生成模型在物理動態與因果一致性上仍有困難,且難以取得可靠的品質判斷與修正配對。
  2. PhysLoop 在生成、選擇與修正各階段引導並約束 LLM,目標是蒸餾物理知識並支援可控資料建構與推論時修復。
  3. 作者稱釋出 100K 支室內外場景影片、生成工具,以及 PhysLoop-Gen 與 PhysLoop-Repair 兩個評測資料集,並在 Physics-IQ Verified 上評估。

閱讀限制

僅依標題與摘要整理,未閱讀全文,無法確認實驗設計、基準細節、資料集品質或同儕審查狀態;所述成果均為作者主張。

原始摘要 [EN]

Video generation models produce visually realistic videos but still struggle with physical dynamics and causal consistency. Learning from generated videos requires fine-grained physical supervision, yet reliable quality judgments, correction pairs, and controlled variations are difficult to obtain. Large language models (LLMs) possess physical knowledge, but struggle to reliably guide video generation models in producing their training videos. We present PhysLoop, a unified harness designed to support physically grounded self-improvement in video generation models. The harness guides and constrains LLMs throughout video generation, selection, and correction, distilling their physical knowledge into physically consistent videos. The framework supports controllable training-data construction and inference-time video repair. By identifying and correcting failures, PhysLoop marks a key milestone toward recursive self-improvement (RSI) in video generation models with limited human intervention. We release 100K videos spanning diverse, realistic indoor and outdoor scenes, along with generation tools and two evaluation datasets: PhysLoop-Gen and PhysLoop-Repair. We evaluate controllable generation and localized repair on these datasets, respectively, and assess physical behavior on the external Physics-IQ Verified benchmark.

AI 摘要僅根據公開標題與摘要整理;研究結論來自作者,請以原文為準。