![[2602.19538] Cost-Aware Diffusion Active Search](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.19538] Cost-Aware Diffusion Active Search
Summary
The paper presents a novel approach to active search using cost-aware diffusion models, improving efficiency in decision-making for autonomous agents in unknown environments.
Why It Matters
This research addresses the challenges of balancing exploration and exploitation in active search scenarios, particularly in robotics and AI. By leveraging diffusion models, it offers a more computationally efficient alternative to traditional search tree methods, which is crucial for real-time applications in dynamic environments.
Key Takeaways
- Introduces a cost-aware diffusion model for active search.
- Improves decision-making efficiency without exhaustive search trees.
- Addresses optimism bias in previous reinforcement learning approaches.
- Demonstrates superior performance in offline reinforcement learning.
- Applicable to both single and multi-agent systems.
Computer Science > Robotics arXiv:2602.19538 (cs) [Submitted on 23 Feb 2026] Title:Cost-Aware Diffusion Active Search Authors:Arundhati Banerjee, Jeff Schneider View a PDF of the paper titled Cost-Aware Diffusion Active Search, by Arundhati Banerjee and Jeff Schneider View PDF HTML (experimental) Abstract:Active search for recovering objects of interest through online, adaptive decision making with autonomous agents requires trading off exploration of unknown environments with exploitation of prior observations in the search space. Prior work has proposed information gain and Thompson sampling based myopic, greedy approaches for agents to actively decide query or search locations when the number of targets is unknown. Decision making algorithms in such partially observable environments have also shown that agents capable of lookahead over a finite horizon outperform myopic policies for active search. Unfortunately, lookahead algorithms typically rely on building a computationally expensive search tree that is simulated and updated based on the agent's observations and a model of the environment dynamics. Instead, in this work, we leverage the sequence modeling abilities of diffusion models to sample lookahead action sequences that balance the exploration-exploitation trade-off for active search without building an exhaustive search tree. We identify the optimism bias in prior diffusion based reinforcement learning approaches when applied to the active search setting and pr...