SealQA: Raising the Bar for Reasoning in Search-Augmented Language Models

TL;DR

SealQA is a new benchmark for evaluating the reasoning and factual accuracy of search-augmented language models in noisy, conflicting web search scenarios, revealing significant limitations of current models.

Contribution

The paper introduces SealQA, a comprehensive benchmark for assessing reasoning in search-augmented models, highlighting their vulnerabilities and setting a new standard for future evaluations.

Findings

Current frontier models perform poorly on SealQA, with accuracy often below 20%.

Increasing compute at test time does not significantly improve model performance.

Models struggle to identify relevant documents in long-context, multi-document settings.

Abstract

We introduce SealQA, a new challenge benchmark for evaluating SEarch-Augmented Language models on fact-seeking questions where web search yields conflicting, noisy, or unhelpful results. SealQA comes in three flavors: (1) Seal-0 (main) and (2) Seal-Hard, which assess factual accuracy and reasoning capabilities, with Seal-0 focusing on the most challenging questions where chat models (e.g., GPT-4.1) typically achieve near-zero accuracy; and (3) LongSeal, which extends SealQA to test long-context, multi-document reasoning in "needle-in-a-haystack" settings. Our evaluation reveals critical limitations in current models: Even frontier LLMs perform poorly across all SealQA flavors. On Seal-0, frontier agentic models equipped with tools like o3 and o4-mini achieve only 17.1% and 6.3% accuracy, respectively, at their best reasoning efforts. We find that advanced reasoning models such as DeepSeek-R1-671B and o3-mini are highly vulnerable to noisy search results. Notably, increasing test-time compute does not yield reliable gains across o3-mini, o4-mini, and o3, with performance often plateauing or even declining early. Additionally, while recent models are less affected by the "lost-in-the-middle" issue, they still fail to reliably identify relevant documents in LongSeal when faced with numerous distractors. To facilitate future work, we release SealQA at huggingface.co/datasets/vtllms/sealqa.