analiza aiqbench
This research reframes interactive navigation as a cost-sensitive uncertainty reduction problem, introducing a new benchmark and metric that penalizes queries based on their informational efficiency. By leveraging information-gain analysis to derive question weights, the proposed zero-shot MLLM navigator selectively interacts only when the expected uncertainty reduction justifies the cost, addressing a critical gap in existing benchmarks that ignore interaction penalties. This work provides a more realistic evaluation framework for embodied agents, shifting the focus from raw success rates to the efficiency of disambiguation strategies.
Ask When It Pays: Cost-Aware Open-Ended Interaction for Instance Goal Navigation
arXiv:2606.03175v1 Announce Type: cross Abstract: Instance Goal Navigation (IGN) requires an embodied agent to find a specific object instance among distractors from an underspecified natural-language description. Such ambiguity often cannot be resolved from perception and language alone, making interaction with an oracle a natural mechanism for disambiguation. Prior interactive methods allow oracle queries but treat lightweight clarification and route-level guidance alike, letting agents boost success rate through repeated high-information questions rather than by resolving the underlying ambiguity efficiently. We recast interactive IGN as a cost-sensitive uncertainty-reduction problem, where the agent should ask the question whose answer provides the largest reduction in navigation uncertainty relative to its penalty. To this end, we apply an information-gain analysis on existing navigation corpora to identify which cues reduce navigation uncertainty, yielding a compact set of question types and data-derived weights.However, existing interactive navigation benchmarks do not model the cost of different question types or evaluate how efficiently agents use interaction, making them unsuitable for studying cost-sensitive interaction. Based on this taxonomy, we construct a benchmark for diagnosing interaction behavior and efficiency, together with a Weighted Success Rate metric that penalizes each query by its derived cost. We further propose a zero-shot MLLM navigator that selectively queries at each decision step only when the expected uncertainty reduction justifies the interaction cost.
Abstract:Instance Goal Navigation (IGN) requires an embodied agent to find a specific object instance among distractors from an under-specified natural-language description. Such ambiguity often cannot be resolved from perception and language alone, making interaction with an oracle a natural mechanism for disambiguation. Prior interactive methods allow oracle queries but treat lightweight clarification and route-level guidance alike, letting agents boost success rate through repeated high-information questions rather than by resolving the underlying ambiguity efficiently. We recast interactive IGN as a cost-sensitive uncertainty-reduction problem, where the agent should ask the question whose answer provides the largest reduction in navigation uncertainty relative to its penalty. To this end, we apply an information-gain analysis on existing navigation corpora to identify which cues reduce navigation uncertainty, yielding a compact set of question types and data-derived weights. However, existing interactive navigation benchmarks do not model the cost of different question types or evaluate how efficiently agents use interaction, making them unsuitable for studying cost-sensitive interaction. Based on this taxonomy, we construct a benchmark for diagnosing interaction behavior and efficiency, together with a Weighted Success Rate metric that penalizes each query by its derived cost. We further propose a zero-shot MLLM navigator that selectively queries at each decision step only when the expected uncertainty reduction justifies the interaction cost.
Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
arXiv:2606.03175 [cs.CV]
(or arXiv:2606.03175v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2606.03175 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 309 cuvinte
Subiecte conexe
Underwriting Superintelligence: Backing Agents you can Sue — Rune Kvist, AIUC
Latent Space · 16 sept. 2026
Ask When It Pays: Cost-Aware Open-Ended Interaction for Instance Goal Navigation
arXiv · 4 iun. 2026
If This Were a Test, How Much Would It Cost?
LessWrong · 16 iun. 2026
How GoFundMe became America’s backup plan
The Verge · 24 aug. 2026