Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

UESF-Bench: Benchmarking and Probing for Unified Embodied Seeking and Following

Source

arXiv

arxiv.org/abs/2607.13621v1

Read the source

Source

Plain-Language Brief

The authors introduce the Unified Embodied Seeking and Following Benchmark (UESF-Bench), a new benchmark for embodied agents that allows them to search for and follow language-described targets in dynamic environments. The authors propose a vision-language-action framework called SeekFollow-VLA, which uses a task-driven routing mechanism to help the agent transition between seeking and following phases.

What the Document Says

  • The UESF-Bench is introduced as a new benchmark for embodied agents. The main goal is to have agents search for and follow language-described targets in dynamic environments.
  • SeekFollow-VLA, a vision-language-action framework, is proposed for the benchmark. It includes a task-driven routing mechanism for latent phase inference and transition modeling between seeking and following phases.
  • The authors claim that SeekFollow-VLA achieves clear improvements over both single-head and dual-head baselines across single-person and multi-person environments.

Key Concepts Mentioned

  • UESF-Bench
  • embodied agents
  • language-guided human following
  • vision-language-action framework
  • SeekFollow-VLA

Claims Made by the Document

  • UESF-Bench addresses the limitations of existing benchmarks by requiring agents to handle semantic-guided exploration, reliable behavior switching, and delayed identity grounding.
  • SeekFollow-VLA achieves clear improvements over both single-head and dual-head baselines across single-person and multi-person environments.

Evidence Used

  • The text describes the UESF-Bench as a benchmark that requires agents to handle specific tasks, such as semantic-guided exploration, reliable behavior switching, and delayed identity grounding.
  • Experimental results are provided showing SeekFollow-VLA's improvements over baselines.

Methods / Approach

  • SeekFollow-VLA uses a task-driven routing mechanism for latent phase inference and transition modeling between seeking and following phases.

Limitations Stated by the Authors

The authors do not mention any limitations of their proposed framework.

Limitations Noted by the Agent

Agent interpretation — not from the source.

It is unknown if there are any limitations to the UESF-Bench or SeekFollow-VLA that were not addressed by the authors.

Open Questions

  • How does SeekFollow-VLA compare to other existing approaches in terms of performance and efficiency?
  • What is the scalability of the proposed framework for larger and more complex environments?

Notes for Future Q&A

May include interpretation. Labeled where it does.

The text focuses on introducing a new benchmark and a vision-language-action framework, but it does not provide detailed comparisons to other existing approaches or information about the framework's scalability.

Raw Abstract / Excerpt

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 a more realistic requirement: an agent often needs to first find a language-described target and then persistently follow that target in a dynamic environment. While recent work has started to study human search, existing settings are typically evaluated in task-specific scenarios and often rely on stronger prior knowledge of the environment. Moreover, they usually treat searching and following as separate tasks and still lack a unified benchmark for systematic evaluation. To address these limitations, we introduce the Unified Embodied Seeking and Following Benchmark (UESF-Bench), a large-scale and diverse benchmark for embodied human seeking and following. The benchmark requires agents to handle semantic-guided exploration, reliable behavior switching and recovery, and delayed identity grounding. To this end, we propose SeekFollow-VLA, a vision-language-action framework with a task-driven routing mechanism for latent phase inference and transition modeling between seeking and following. Experimental results show that SeekFollow-VLA achieves clear improvements over both single-head and dual-head baselines across single-person and multi-person environments, establishing a baseline for unified embodied seek-and-follow.