Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper analyzes the limitations of current AI benchmarks in reflecting real-world performance in low-resource contexts, highlighting discrepancies between lab evaluations and deployment conditions. It advocates for evaluating deployed systems holistically, considering factors like noisy inputs, limited connectivity, and low-end hardware, rather than focusing solely on isolated model performance. The authors propose a shared reporting framework with standardized benchmark cards and deployment profiles to improve comparability and inform decision-making for policymakers and implementers.
Current AI benchmarks are misleading because they ignore the messy reality of low-resource deployments, where system-level constraints matter as much as model accuracy.
Existing AI evaluation practices often fail to capture how systems actually perform in low-resource environments, where operational constraints shape usability as much as model quality. Through a structured analysis of existing benchmark families across speech, chat/RAG, and vision systems, we identify critical gaps between laboratory evaluation practices and real-world deployment conditions in low-resource environments. We argue that the meaningful unit of assessment is the deployed system rather than an isolated model and that effective evaluation frameworks must integrate task performance with deployment conditions such as noisy inputs, code-switching, intermittent connectivity, low-end hardware, and domain shift. At the same time, benchmarks should recognize that different application classes require distinct evaluation profiles rather than a single aggregate score that obscures operational differences. To support practical decision-making, we propose a shared reporting framework that preserves comparability across systems and application types while remaining sensitive to deployment context. Finally, we emphasize the need for concise and actionable reporting artifacts for policymakers, donors, and implementers, including standardized one-page benchmark cards, deployment profiles, and explicit documentation of failure handling procedures and human oversight mechanisms.