Despite strong performance on existing benchmarks, it remains unclear whether large language models can reason over genuinely novel scientific information. Most evaluations score end-to-end RAG pipelines, where reasoning is confounded with retrieval and toolchain choices, and the signal is further contaminated by parametric memorization and open-web volatility. We introduce DeR2, a controlled deep-research sandbox that isolates document-grounded reasoning while preserving core difficulties of deep search: multi-step synthesis, denoising, and evidence-based conclusion making. DeR2 decouples evidence access from reasoning via four regimes--Instruction-only, Concepts (gold concepts without documents), Related-only (only relevant documents), and Full-set (relevant documents plus topically related distractors)--yielding interpretable regime gaps that operationalize retrieval loss vs. reasoning loss and enable fine-grained error attribution. To prevent parametric leakage, we apply a two-phase validation that requires parametric failure without evidence while ensuring oracle-concept solvability. To ensure reproducibility, each instance provides a frozen document library (drawn from 2023-2025 theoretical papers) with expert-annotated concepts and validated rationales. Experiments across a diverse set of state-of-the-art foundation models reveal substantial variation and significant headroom: some models exhibit mode-switch fragility, performing worse with the Full-set than with Instruction-only, while others show structural concept misuse, correctly naming concepts but failing to execute them as procedures.

翻译：尽管现有基准测试表现出色，但大型语言模型是否能够对真正新颖的科学信息进行推理仍不明确。大多数评估方法对端到端RAG流水线进行评分，其中推理过程与检索及工具链选择相互混淆，且评估信号进一步受到参数记忆和开放网络波动性的干扰。我们提出DeR2——一个受控的深度研究沙盒，该框架在保持深度搜索核心难点（多步综合、去噪和基于证据的结论推导）的同时，实现了文档驱动推理的隔离评估。DeR2通过四种机制解耦证据获取与推理过程：仅指令模式、概念模式（无文档的黄金概念）、相关文档模式（仅相关文档）和全集模式（相关文档加主题相关干扰项），由此产生的可解释机制差异将检索损失与推理损失操作化，支持细粒度错误归因。为防止参数泄露，我们采用两阶段验证机制：要求模型在无证据时必然失败，同时确保基于预言概念的可行性。为保证可复现性，每个测试实例均提供冻结文档库（源自2023-2025年理论论文）及专家标注的概念与验证推理链。对多种前沿基础模型的实验表明存在显著性能差异与提升空间：部分模型呈现模式切换脆弱性，在全集模式下的表现反而差于仅指令模式；另一些模型则表现出结构性概念误用，能够正确指称概念但无法将其作为操作流程执行。