Search papers, labs, and topics across Lattice.
The authors introduce DR$^{3}$-Eval, a benchmark for evaluating deep research agents on multimodal, multi-file report generation using authentic user-provided materials and a static research sandbox corpus. This benchmark addresses the challenges of dynamic web environments and ambiguous task definitions in evaluating DRAs. Experiments using their DR$^{3}$-Agent reveal critical failure modes in retrieval robustness and hallucination control, highlighting the benchmark's difficulty.
Current research agents still struggle with retrieval robustness and hallucination control, even when evaluated in a static, verifiable research environment.
Deep Research Agents (DRAs) aim to solve complex, long-horizon research tasks involving planning, retrieval, multimodal understanding, and report generation, yet their evaluation remains challenging due to dynamic web environments and ambiguous task definitions. We propose DR$^{3}$-Eval, a realistic and reproducible benchmark for evaluating deep research agents on multimodal, multi-file report generation. DR$^{3}$-Eval is constructed from authentic user-provided materials and paired with a per-task static research sandbox corpus that simulates open-web complexity while remaining fully verifiable, containing supportive documents, distractors, and noise. Moreover, we introduce a multi-dimensional evaluation framework measuring Information Recall, Factual Accuracy, Citation Coverage, Instruction Following, and Depth Quality, and validate its alignment with human judgments. Experiments with our developed multi-agent system DR$^{3}$-Agent based on multiple state-of-the-art language models demonstrate that DR$^{3}$-Eval is highly challenging and reveals critical failure modes in retrieval robustness and hallucination control. Our code and data are publicly available.
