Summary of Step: Enhancing Video-llms’ Compositional Reasoning by Spatio-temporal Graph-guided Self-training, By Haiyi Qiu et al.
STEP: Enhancing Video-LLMs’ Compositional Reasoning by Spatio-Temporal Graph-guided Self-Training
by Haiyi Qiu, Minghe Gao, Long Qian, Kaihang Pan, Qifan Yu, Juncheng Li, Wenjie Wang, Siliang Tang, Yueting Zhuang, Tat-Seng Chua
First submitted to arxiv on: 29 Nov 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A recent surge in Video Large Language Model (Video-LLM) capabilities has led to impressive results in basic video understanding tasks. However, these models struggle with complex multi-step spatio-temporal inference required for compositional reasoning. To overcome this limitation, the authors propose STEP, a novel graph-guided self-training method that enables Video-LLMs to generate fine-tuning data from raw videos and improve their reasoning abilities. The approach involves inducing Spatio-Temporal Scene Graph (STSG) representations of diverse videos, deriving multi-step reasoning Question-Answer (QA) data with Chain-of-Thought (CoT) rationales, and integrating answers and rationales as a training objective. Experimental results demonstrate the effectiveness of STEP across models of varying scales, achieving significant improvements in tasks requiring multiple reasoning steps. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Video Large Language Models are getting better at understanding videos! But they’re not great at figuring out complex things that happen over time. To help them get better, researchers created a new way to train these models using graphs and chains of thoughts. They showed that this method can make the models much better at answering questions and understanding what’s happening in videos. |
Keywords
» Artificial intelligence » Fine tuning » Inference » Large language model » Self training
