Baikal: Structured Search for Deep Research over Data Lakes

2026-07-30

AI Summary

This study addresses the limitation of existing deep data lake exploration methods, which often over-exploit local regions and fail to adequately cover diverse semantic areas under a fixed budget, resulting in reports lacking comprehensiveness and diversity. The authors formulate this challenge as a budget-constrained semantic search problem and propose a structured semantic exploration framework. This framework first constructs semantic regions via heterogeneous data clustering and then adaptively balances exploration and exploitation across regions. Subproblems are generated per region, and large language model (LLM)-driven retrieval and reasoning assess the quality of findings. Region values are dynamically updated using reinforcement learning strategies—including Bayesian ε-greedy, UCB, and LLM-guided approaches—to guide subsequent search steps. Experiments on HybridQA and TAT-QA demonstrate that the proposed method improves report scores by 28% and 36% over the strongest baseline, significantly enhancing coverage, relevance, verifiability, and practical utility.

Abstract

Deep research over data lakes requires an LLM agent to investigate evidence across thousands of heterogeneous tables and passages to synthesize a report. Existing methods perform iterative retrieval and generation, letting accumulated context determine what to investigate next, which can overexploit locally promising evidence and fail to cover distinct semantic regions under a fixed budget. To address this, we cast deep research over data lakes as a budgeted search problem and present Baikal - a framework that clusters heterogeneous evidence into semantic regions, then searches over them adaptively to balance exploration and exploitation. Within each selected region, Baikal generates and investigates region-grounded subquestions, using finding quality as rewards to update region-level value estimates and guide search under policies ranging from random and LLM-guided selection to Bayesian $ε$-greedy and UCB. We evaluate Baikal on 15 queries each over HybridQA and TAT-QA data lakes containing 10,993 and 2,757 tables, respectively, together with 227K Wikipedia passages and 13K financial report passages. We assess research quality with a new rubric covering groundedness, relevance, diversity, and utility, and use GPT-5-mini to score Baikal and strong baselines, including DeepSearcher and an OpenCode research agent with retrieval and clustering variants. Across both data lakes, Baikal performs strongly under several region-selection policies; its best configuration improves report scores over the strongest baselines by 28% on HybridQA and 36% on TAT-QA. Our analyses attribute these gains to organizing and exploring semantic evidence regions, which improves groundedness and diversity and yields more useful findings under the same subquestion budget. These results demonstrate the value of structured semantic exploration for systematic research and discovery over heterogeneous data lakes.

Problem

• data lakes
• deep research
• semantic exploration
• evidence coverage
• budgeted search

Innovation

• structured search
• semantic clustering
• budgeted exploration
• LLM agent
• data lakes