Search papers, labs, and topics across Lattice.
The paper introduces MERRIN, a new benchmark designed to evaluate search-augmented agents' ability to retrieve and reason over multimodal evidence from noisy web environments using natural language queries. The benchmark incorporates diverse modalities like video and audio and requires multi-hop reasoning across conflicting web sources. Experiments using ten models, including GPT-4-mini and Gemini, reveal that even the strongest agents struggle, achieving only 40.1% accuracy, due to over-exploration and inefficient source selection, highlighting the need for more robust multimodal reasoning capabilities.
Even the best search-augmented agents, like Gemini Deep Research, are easily distracted by noisy web content, leading to surprisingly poor performance (40% accuracy) on a new multimodal reasoning benchmark.
Motivated by the underspecified, multi-hop nature of search queries and the multimodal, heterogeneous, and often conflicting nature of real-world web results, we introduce MERRIN (Multimodal Evidence Retrieval and Reasoning in Noisy Web Environments), a human-annotated benchmark for evaluating search-augmented agents. MERRIN measures AI agents'ability to identify relevant modalities, retrieve multimodal evidence, and perform multi-hop reasoning over noisy web sources. It differs from prior work in three important aspects: (1) using natural language queries without explicit modality cues, (2) incorporating underexplored modalities such as video and audio, and (3) requiring the retrieval of complex, often noisy or conflicting multimodal evidence during web search. We evaluate diverse search agents powered by ten models, including strong closed-source models (e.g., GPT-5.4-mini, Gemini 3/3.1 Flash/Pro) and open-weight models (Qwen3-4B/30B/235B), across three search settings (no search, native search, and agentic search). Our results show that MERRIN is highly challenging: the average accuracy across all agents is 22.3%, with the best-performing agent reaching only 40.1%. We further observe that while stronger agents like Gemini Deep Research achieve higher performance, gains are modest due to over-exploration; they take more steps and use more tools, but are often distracted by conflicting or partially relevant web content, leading to incorrect answers. Compared to humans, these agents consume more resources yet achieve lower accuracy, largely due to inefficient source selection and an overreliance on text modalities. These findings highlight the need for search agents capable of robust search and reasoning across diverse modalities in noisy web environments, making MERRIN a valuable testbed for evaluating such capabilities.
