BaRA: Budget-constrained and Reliable Web Data Collection Agent
About
Large language model (LLM)-based web agents automate web navigation and data collection. However, live web data collection demands capabilities beyond task completion: agents must discover site-internal pages and retrieve text, image, and video artifacts in an accessible form within a fixed interaction budget. We formulate this setting as budget-constrained, site-level multimodal web data collection and propose Budget-constrained and Reliable Agent (BaRA). BaRA performs breadth-first search (BFS)-based link discovery with liveness verification to filter hallucinated and dead links, then validates extracted multimodal artifacts using rule-based provenance and accessibility checks. A history-based self-reflection module recovers from execution failures and incomplete outputs. On controlled synthetic and real-world websites, BaRA consistently improves valid-link discovery and download-valid multimodal extraction over existing agents. Our code is available at https://github.com/MLAI-Yonsei/BaRA-Agent.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Link discovery | loc.gov Internal Link Discovery (Step 1) | Precision100 | 4 | |
| Multimodal Text Extraction | Synthetic | Precision96.07 | 4 | |
| Multimodal Video Extraction | Synthetic | Precision100 | 4 | |
| Multimodal Video Extraction | LOC | Precision100 | 4 | |
| Video Extraction | Synthetic oracle-link setting | Precision100 | 4 | |
| Video Extraction | LOC oracle-link | Precision100 | 4 | |
| Image Extraction | Synthetic oracle-link setting | Precision0.9769 | 4 | |
| Image Extraction | LOC oracle-link setting | Precision88.34 | 4 | |
| Link discovery | Synthetic Internal Link Discovery (Step 1) | Precision100 | 4 | |
| Multimodal Image Extraction | Synthetic | Precision97.69 | 4 |