MERGE: Guided Vision-Language Models for Multi-Actor Event Reasoning and Grounding in Human�Robot Interaction
Joerg Deigmoeller, Nakul Agarwal, Stephan Hasler, Daniel Tanneberg, Anna Belardinelli, Reza Ghoddoosian, Chao Wang, Felix Ocker, Fan Zhang, Behzad Dariush, Michael Gienger
AI summary
Problem
Current vision-language models struggle with persistent instance tracking and temporal consistency in dynamic multi-actor settings, while existing benchmarks lack fine-grained, role-aware annotations for collaborative human-robot interactions.
Approach
MERGE uses a lightweight streaming perception pipeline to continuously track actors and objects, triggering a Vision-Language Model only during salient changes to generate temporally consistent actor-action-object event tuples.
Key results
- Introduces MERGE framework for persistent multi-actor event grounding
- Releases GROUND dataset with fine-grained role-aware annotations for collaborative HRI
- Achieves 2x higher average grounding score than GPT-4o, GPT-5, and Gemini 2.5 Flash
- Reduces inference runtime by 4x compared to continuous frame-by-frame VLM captioning
Why it matters
Provides a scalable, efficient foundation for robots to maintain real-time situational awareness and understand complex collaborative interactions in group settings.
Abstract
We introduce MERGE, a system for situational grounding of actors, objects, and events in dynamic hu- man–robot group interactions. Effective collaboration in such settings requires consistent situational awareness, built on persistent representations of people and objects and an episodic abstraction of events. MERGE achieves this by uniquely iden- tifying physical instances of actors (humans or robots) and objects and structuring them into actor–action–object relations, ensuring temporal consistency across interactions. Central to MERGE is the integration of Vision-Language Models (VLMs) guided with a perception pipeline: a lightweight streaming module continuously processes visual input to detect changes and selectively invokes the VLM only when necessary. This decoupled design preserves the reasoning power and zero-shot generalization of VLMs while improving efficiency, avoiding both the high monetary cost and the latency of frame-by- frame captioning that leads to fragmented and delayed outputs. To address the absence of suitable benchmarks for multi- actor collaboration, we introduce the GROUND dataset, which offers fine-grained situational annotations of multi-person and human–robot interactions. On this dataset, our approach im- proves the average grounding score by a factor of 2 compared to the performance of VLM-only baselines–including GPT- 4o, GPT-5 and Gemini 2.5 Flash–while also reducing run- time by a factor of 4. The code and data are available at www.github.com/HRI-EU/merge.