New Benchmark for Embodied AI Human Seeking and Following

Kun Yu, Jianhua Yang, Yixiang Chen, Changwei Wang, Hongyuan Yu, Yan Huang, Fushuo Huo, Ya Jing, Zhumin Chen, Keji He· July 16, 2026 View original

Summary

Researchers introduce UESF-Bench, a new large-scale benchmark for embodied agents to first find a language-described human target and then persistently follow them in dynamic environments, addressing limitations of prior task-specific scenarios.

A new research paper presents UESF-Bench, a novel benchmark designed to evaluate embodied AI agents in more realistic human interaction scenarios. Unlike previous benchmarks that often assume the target is immediately visible, UESF-Bench challenges agents to actively search for a language-described person before initiating a following behavior. This unified approach integrates semantic-guided exploration, reliable behavior switching, and delayed identity grounding, reflecting the complexities of real-world dynamic environments. To tackle this, the paper also proposes SeekFollow-VLA, a vision-language-action framework that uses a task-driven routing mechanism to manage transitions between seeking and following phases. Experimental results demonstrate that SeekFollow-VLA significantly outperforms existing baselines in both single and multi-person settings, establishing a new foundation for future research in this area.

Why it matters

This research is crucial for advancing the capabilities of embodied AI, enabling robots and autonomous systems to interact more naturally and effectively with humans in complex, real-world settings.

How to implement this in your domain

  1. 1Review the UESF-Bench framework for potential application in developing more robust human-robot interaction systems.
  2. 2Explore the SeekFollow-VLA architecture to inform the design of vision-language-action models for autonomous agents.
  3. 3Consider integrating semantic-guided exploration and dynamic behavior switching into existing robotic navigation or assistance platforms.
  4. 4Participate in the benchmark to test and validate new algorithms for embodied seeking and following tasks.

Who benefits

RoboticsLogisticsHealthcareSmart HomesDefense

Key takeaways

  • UESF-Bench offers a unified benchmark for embodied AI to seek and follow humans.
  • The benchmark addresses realistic scenarios requiring initial search and persistent following.
  • SeekFollow-VLA framework improves performance in dynamic, multi-person environments.
  • This research advances human-robot interaction and autonomous system capabilities.

Original post by Kun Yu, Jianhua Yang, Yixiang Chen, Changwei Wang, Hongyuan Yu, Yan Huang, Fushuo Huo, Ya Jing, Zhumin Chen, Keji He

"arXiv:2607.13621v1 Announce Type: new Abstract: Language-guided human following is an important capability for embodied agents, but existing benchmarks typically assume that the target person is visible at the start of an episode. This setting simplifies the problem and overlooks…"

View on X

Originally posted by Kun Yu, Jianhua Yang, Yixiang Chen, Changwei Wang, Hongyuan Yu, Yan Huang, Fushuo Huo, Ya Jing, Zhumin Chen, Keji He on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses

More in AI Research

AI Engineering & DevToolsAI Research

NodeImport Improves Imbalanced Node Classification on Graphs

NodeImport is a new framework addressing class imbalance in graph node classification by assessing node importance to create a balanced meta-set for training. It dynamically filters valuable labeled, unlabeled, and synthetic nodes, outperforming existing baselines across various datasets and GNN architectures.

Nan Chen, Zemin Liu, Bryan Hooi, Bingsheng He, Jun Hu, Jia ChenJul 16, 2026
AI ResearchAI Engineering & DevTools

Neural Spline Flows Aid Dark Matter Search in CMS Data.

This paper reports a search for dark matter produced with a leptonically decaying Z boson using CMS Run 2015D open data and Neural Spline Flows. The method models signal and background densities to set upper limits on signal-strength parameters for various dark matter mediators, though observed limits are weaker than expected due to background modeling discrepancies.

Hitesh Rasineni (VIT-AP University, Amaravati, India), Bhavishya Chebrolu (Mohan Babu University, Tirupati, India)Jul 16, 2026
AI Engineering & DevToolsAI Research

Multiplex Graph Transformer Boosts Power Grid Model Generalization.

Researchers introduce MxGPS, a multiplex graph transformer designed to overcome "topology overfitting" in power grid problems. By jointly training on multiple tasks with a shared encoder, MxGPS achieves superior zero-shot generalization across unseen grid topologies, demonstrating high accuracy and low boundary violation rates with significantly fewer parameters.

Charilaos Papaioannou, Ioannis Tsantilas, Dimitris Giannakakos, Vasilis Michalakopoulos, Sotiris Pelekis, Vangelis Marinakis, Arsam Aryandoust, Antonello Monti, Ricardo J. Bessa, Perdo P. Vergara, Jochen Cremer, Elissaios SarmasJul 16, 2026