Xi'an Jiaotong's QQWorld Boosts World Model Planning Success Rate
机器之心 · wechat · 2026-08-11
To address the heavy-tail issue in the latent space of Yann LeCun's LeWorldModel (LeWM), researchers from Xi'an Jiaotong University proposed QQWorld. They found that the EP normality objective used in LeWM has a 'tail blind spot,' struggling to correct extreme latent variables that drift far from the main distribution.
QQWorld replaces the EP regularizer with Quantile-Quantile Matching, directly moving each projected sample to its corresponding standard Gaussian quantile. This provides continuous effective gradients at the distribution tails, with a core implementation of less than 10 lines of code.
Experiments show QQWorld increases the average planning success rate from 79.75% to 85.08% across multiple control environments. Additionally, the proposed Cross-Batch QQ mechanism uses historical batch features to stabilize ordering, achieving significantly better performance than the baseline while reducing physical batch size to a quarter and cutting GPU memory by 73%.
More from Research
- Dyna Robotics' New Model Trained on 1M Hours of Human Video Hits 90% Task Success — Dr_Singularity · 2026-08-11
- Exploring Knowledge Cutoffs and Pre-training Timelines for Claude & GPT — sshh12 · 2026-08-11
- Sakana AI's CEDAR uses LLM agents to design complex system dynamics — dair_ai · 2026-08-11
- Why Deep Networks Escape the Curse of Dimensionality: Matthieu Wyart — Machine Learning Street Talk · 2026-08-11
- MIT Tech Review Explores the Next Era of LLMs: Startups Challenging the Transformer — volokuleshov · 2026-08-11
- Thrive Researcher Shares Talks on Frontier Context Engineering and Model Interpretability — thesephist · 2026-08-11