Complex, multi-step reasoning demands agents to be capable of dynamically exploring diverse research directions and adapting plans on the fly. We introduce Gnosis Deepsearch, a schema-guided framework for long-horizon planning in complex Question Answering, where variables, clues, and constraints are captured in a structured state representation maintained across trajectories. This state is cascaded into parallel search paths broadening exploration without relying on sequential execution, thereby mitigating both limited search space coverage and catastrophic forgetting. We couple Gnosis with a retriever that sidesteps challenges associated with retrieval and LLM reading when longer documents are involved. Our retrieval strategy ensures that the most relevant snippets from longer documents are considered, showcasing several benefits within both Gnosis and a more general ReAct framework on the BrowseComp-Plus benchmark. Extensive experiments on several QA benchmarks of varying difficulty demonstrate the strong performance of Gnosis across different configurations.

Real-world financial analysis involves information across multiple languages and modalities, from reports and news to scanned filings and meeting recordings. Yet most existing evaluations of LLMs in finance remain text-only, monolingual, and largely saturated by current models. To bridge these gaps, we present MultiFinBen, the first expert-annotated multilingual (five languages) and multimodal (text, vision, audio) benchmark for evaluating LLMs in realistic financial contexts. MultiFinBen introduces two new task families: multilingual financial reasoning, which tests cross-lingual evidence integration from filings and news, and financial OCR, which extracts structured text from scanned documents containing tables and charts. Rather than aggregating all available datasets, we apply a structured, difficulty-aware selection based on advanced model performance, ensuring balanced challenge and removing redundant tasks. Evaluating 21 leading LLMs shows that even frontier multimodal models like GPT-4o achieve only 46.01% overall, stronger on vision and audio but dropping sharply in multilingual settings. These findings expose persistent limitations in multilingual, multimodal, and expert-level financial reasoning. All datasets, evaluation scripts, and leaderboards are publicly released.

Retrieval-augmented Generation (RAG) relies on effective retrieval capabilities, yet traditional sparse and dense retrievers inherently struggle with multi-hop retrieval scenarios. In this paper, we introduce GeAR, a system that advances RAG performance through two key innovations: (i) an efficient graph expansion mechanism that augments any conventional base retriever, such as BM25, and (ii) an agent framework that incorporates the resulting graph-based retrieval into a multi-step retrieval framework. Our evaluation demonstrates GeAR’s superior retrieval capabilities across three multi-hop question answering datasets. Notably, our system achieves state-of-the-art results with improvements exceeding 10% on the challenging MuSiQue dataset, while consuming fewer tokens and requiring fewer iterations than existing multi-step retrieval systems. The project page is available at https://gear-rag.github.io.