Search papers, labs, and topics across Lattice.
This paper introduces GauntletBench, a novel benchmark designed to evaluate the generalization capabilities of agentic systems in complex, real-world scenarios, focusing on underexplored dimensions such as temporal perception, graphical understanding, and 3D reasoning. The benchmark includes five professional applications with a total of 100 vision-intensive tasks and provides a modular evaluation pipeline compatible with various agent frameworks. Empirical results indicate that even state-of-the-art agents only achieve a 19.1% success rate, starkly contrasting with over 80% success from non-expert human annotators, underscoring significant limitations in current agent capabilities.
Despite advances in agentic systems, they still lag far behind human performance in complex tasks, achieving only 19.1% success on a new benchmark designed to probe their limitations.
As agentic systems continue to evolve and are widely deployed in real-world scenarios, there is a growing demand to faithfully evaluate their capabilities. However, current benchmarks are typically built on popular applications with relatively simple tasks and focus on a narrow set of capabilities while overlooking broader dimensions, resulting in saturated performance on modern agents and failing to probe their limitations. To this end, we introduce GauntletBench, a web-based benchmark for evaluating agent generalisation in challenging scenarios, focusing on three underexplored capabilities (temporal perception, graphical understanding, and 3D reasoning), across five less-covered professional applications (Video Editor, Workflow Builder, 3D Modeller, Flight Analyser, and Circuit Designer), each with 20 vision-intensive tasks (100 in total). Our benchmark provides a modular pipeline that comprises an environment compatible with both open- and closed-source agent frameworks, a controlled web-based application, a well-structured task suite, and an automated evaluation engine with diverse metrics. Contrary to widespread expectations, our empirical results reveal that frontier agentic systems remain far from achieving human-level performance. Even the state-of-the-art agent achieves only a 19.1% success rate on our GauntletBench, highlighting the limitations in these overlooked capabilities and generalisation. By comparison, non-expert human annotators achieve over 80% success on our challenging yet feasible tasks, revealing the substantial gap between current agent capabilities and those required for complex real-world scenarios.
