FAST-EQA: Efficient Embodied Question Answering with Global and Local Region Relevancy

Haochen Zhang; Nirav Savaliya; Faizan Siddiqui; Enna Sachdeva

cs.ROFeb 17, 2026

FAST-EQA: Efficient Embodied Question Answering with Global and Local Region Relevancy

Authors: Haochen Zhang, Nirav Savaliya, Faizan Siddiqui, Enna Sachdeva

Organizations: Carnegie Mellon University, USA · Honda Research Institute USA

Abstract

Embodied Question Answering (EQA) combines visual scene understanding, goal-directed exploration, spatial and temporal reasoning under partial observability. A central challenge is to confine physical search to question-relevant subspaces while maintaining a compact, actionable memory of observations. Furthermore, for real-world deployment, fast inference time during exploration is crucial. We introduce FAST-EQA, a question-conditioned framework that (i) identifies likely visual targets, (ii) scores global regions of interest to guide navigation, and (iii) employs Chain-of-Thought (CoT) reasoning over visual memory to answer confidently. FAST-EQA maintains a bounded scene memory that stores a fixed-capacity set of region-target hypotheses and updates them online, enabling robust handling of both single and multi-target questions without unbounded growth. To expand coverage efficiently, a global exploration policy treats narrow openings and doors as high-value frontiers, complementing local target seeking with minimal computation. Together, these components focus the agent's attention, improve scene coverage, and improve answer reliability while running substantially faster than prior approaches. On HMEQA and EXPRESS-Bench, FAST-EQA achieves state-of-the-art performance, while performing competitively on OpenEQA and MT-HM3D.

Figures & tables

Figure 2: FAST-EQA processes the question (Q) by extracting relevant regions (R) and visual targets (T). At each step, it localizes its current region ( $\mathrm{R}_{\mathrm{t}}$ ) and updates a semantic memory. For each target $T_{m}$ , it maintains a dedicated memory $\psi_{m}$ that is refined using a visual relevance score. The agent employs a semantic frontier–guided global exploration strategy, leveraging narrow passages (e.g., doors and hallways) to effectively search for relevant semantic regions. When a relevant region is reached, it switches to local exploration to refine the target-specific memory. Once the stopping criterion is satisfied, the agent queries a VLM to generate the final answer. Dotted lines indicate module inputs while solid lines indicate procedural direction of the system.

Figure 3: FAST-EQA leverages language-aligned features from AM-RADIO together with a SigLIP adaptor, to direct the agent toward the target regions. For queries such as (a) Bathroom and (b) Bedroom, the predicted heatmaps are thresholded to produce white contour segments, while the red dot indicates the contour centroid to step towards. This visualization illustrates how semantic grounding enables precise localization of task-relevant areas in the environment to guide exploration from global to local.

Figure 4: FAST-EQA employs a bounded memory system that allocates a dedicated visual memory for each target, retaining the $k$ most relevant images (here, $k=3$ ). The overall memory footprint scales only with the number of targets and remains constant over time, even in long-horizon tasks.

Figure 5: An example from EXPRESS-Bench illustrating how FAST-EQA identifies the relevant region $R$ and target $T$ from the question $Q$ . It then explores the scene, and once the stopping condition is met, correctly generates the final answer from the retrieved visual memory.

	HM-EQA		MT-HM3D		EXPRESS-Bench		A-EQA (184)
	SR	Steps ( $\downarrow$ )	SR	Steps ( $\downarrow$ )	LLM Score	$E_{path}$ ( $\uparrow$ )	LLM-Match	$E_{path}$ ( $\uparrow$ )
GPT-4V (OpenEQA) [ 22 ]	-	-	-	-	-	-	41.8	7.5
Explore-EQA [ 25 ]	58.4	0.52	36.2*	0.64	-	-	46.9*	23.4
Graph-EQA [ 26 ]	63.5	0.20	45.63*	0.45	-	-	30.1* †	-
Memory-EQA [ 37 ]	63.4	0.40	55.1	0.41	-	-	36.8 †	-
Fine-EQA [ 17 ]	56.0	0.54	-	-	63.95	25.58	43.3 †	29.2

Table 1: Performance comparison across EQA benchmarks against SOTA baseline methods. * indicates that the result is from a reproduced experiment reported by others. † indicates results are on full A-EQA split.

	Average Step Time (s)
Fine-EQA	$2.94\pm 0.14$
Explore-EQA	$4.90\pm 0.26$
3D-Mem	$14.53\pm 0.25$
Memory-EQA	$15.89\pm 0.65$
FAST-EQA (Ours)	$\mathbf{2.54\pm 0.05}$

Table 2: Average step time, in seconds, of FAST-EQA compared to other methods on a subset of questions from HM-EQA

	SR	Steps ( $\downarrow$ )
FAST-EQA w/o Doorway Frontiers	67.0	0.64
FAST-EQA w/o CLIP Scoring	69.0	0.63
FAST-EQA w/o VLM Scoring	72.0	0.63
FAST-EQA w/o CoT Reasoning	72.0	0.60
FAST-EQA (Ours)	76.0	0.65

Table 3: Ablation of various components of our EQA system on a fixed subset of HM-EQA

Embodied QA

Embodied question answering studies how agents use perception, memory, and action to answer questions grounded in the physical world.

See all 39 papers

Hierarchical Floorplan-Guided Vision-Language Exploration for Embodied Question Answering
Sep 22, 2026Albert Gassol Puigjaner, Kostas AlexisEmbodied Navigation Embodied QA
Autonomous Robots Lab, Norwegian University of Science and Technology (NTNU), Trondheim, Norway
ScoutVLA: UAV-Centric Active Perception via a Dual-Expert VLA Model for Open-World Embodied Question Answering
Jun 9, 2026Wenhao Lu, Zhengqiu Zhu, Xiaofeng Wang +7Aerial Robotics Embodied QA
1National Key Laboratory of Digital Intelligent Modeling and Simulation, National University of Defense Technology · 2GigaAI
Informative Viewpoint Selection for Episodic-Memory Embodied Question Answering using Omnidirectional Images
Sep 27, 2026Kaname Kitamura, Asako KanezakiEmbodied QA Open-Domain QA
Department of Computer Science, Institute of Science Tokyo, Tokyo, Japan · RIKEN, Japan · Tohoku University, Sendai, Japan

Sep 22, 2026J/K move · Enter open · S save

Hierarchical Floorplan-Guided Vision-Language Exploration for Embodied Question Answering

Albert Gassol Puigjaner, Kostas Alexis

Autonomous Robots Lab, Norwegian University of Science and Technology (NTNU), Trondheim, Norway

Embodied Navigation Embodied QA VLM Reasoning

Open paper AskPDF arXiv

Embodied Question Answering (EQA) requires an agent to explore a previously unseen environment, gather relevant information, and answer questions about the scene. Recent approaches leverage Vision-Language Models (VLMs) together with semantic maps or scene graphs to guide exploration. However, exploration is typically driven only by local observations, while structural priors about the environment remain largely unused. We propose HFLEX-EQA, a hierarchical EQA framework that combines online scene graph construction, VLM- based planning, semantic frontier exploration, and floorplan priors. The system incrementally builds a hierarchical scene graph and an open-vocabulary occupancy map from RGB-D observations, enabling a VLM to jointly reason over the scene graph, task-relevant visual observations, exploration history, and an estimated topological floorplan. Furthermore, we introduce a room-discovery strategy that leverages the floorplan and open-vocabulary frontier semantics to guide exploration toward semantically relevant yet currently unobserved room types. We evaluate HFLEX-EQA on the OpenEQA and ExploreEQA benchmarks and demonstrate deployment on a quadruped robot in real indoor environments. Our results demonstrate the benefit of combining VLM-based hierarchical planning with structural floorplan priors for the EQA task.

Abstract

Embodied QA

Explore similar work

Hierarchical Floorplan-Guided Vision-Language Exploration for Embodied Question Answering

ScoutVLA: UAV-Centric Active Perception via a Dual-Expert VLA Model for Open-World Embodied Question Answering

Informative Viewpoint Selection for Episodic-Memory Embodied Question Answering using Omnidirectional Images