Life After Benchmark Saturation: A Case Study of CORE-Bench
For researchers evaluating AI agents, this work provides a framework for more rigorous evaluation beyond accuracy saturation, though it is a case study on a specific benchmark.
This paper argues that when benchmark accuracy saturates, evaluating agents on dimensions like construct validity, generalizability, efficiency, reliability, model vs. scaffold importance, and human-agent collaboration yields meaningful insights. Using CORE-Bench Hard, they found that even after accuracy saturation, these metrics remain useful, and a human-agent collaboration experiment showed a statistically significant speedup of about 2x.
When a benchmark's accuracy saturates, it is often retired and replaced with a more challenging version. We show that this approach privileges accuracy and misses the opportunity to study six other key dimensions of agent performance: construct validity issues such as shortcuts, out-of-distribution generalizability, efficiency, reliability, the relative importance of the model versus the scaffold, and uplift from human-agent collaboration. We use CORE-Bench Hard, a benchmark for computational reproducibility of scientific code, as a case study to demonstrate that measuring agents along these dimensions yields meaningful insights into agent performance even after accuracy saturates. First, we surface threats to construct validity in CORE-Bench Hard that are difficult to anticipate with less capable agents. We introduce an improved benchmark, CORE-Bench v1.1, and an out-of-distribution task suite, CORE-Bench OOD. Second, we find that despite accuracy saturation, CORE-Bench v1.1 remains useful for measuring efficiency, reliability, model performance, and scaffold performance. Finally, we conduct a small-scale randomized experiment to measure uplift from human-agent collaboration on real-world computational reproducibility tasks. We find a statistically significant speedup by about a factor of two -- likely underestimated due to one-fifth of human-only reproductions reaching the time limit before completing -- and describe various other findings. Together, our contributions present a more rigorous alternative to the dominant accuracy-centric evaluation paradigm.