Preprint · 2026

Life After Benchmark Saturation: A Case Study of CORE-Bench

Nitya Nadgir, Sayash Kapoor, Kangheng Liu, Peter Kirgis, Matilda Orona, Stephan Rabanser, Tilman Bayer, Abhishek Shetty, Yue Ling, Derrick Chan-Sew, Rumi Nakagawa, Saiteja Utpala, Zachary S. Siegel, Arvind Narayanan

Abstract

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.

BibTeX Copy

@misc{nadgir_life_2026,
	title = {Life {After} {Benchmark} {Saturation}: {A} {Case} {Study} of {CORE}-{Bench}},
	shorttitle = {Life {After} {Benchmark} {Saturation}},
	url = {http://arxiv.org/abs/2606.26158},
	doi = {10.48550/arXiv.2606.26158},
	abstract = {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.},
	urldate = {2026-11-06},
	publisher = {arXiv},
	author = {Nadgir, Nitya and Kapoor, Sayash and Liu, Kangheng and Kirgis, Peter and Orona, Matilda and Rabanser, Stephan and Bayer, Tilman and Shetty, Abhishek and Ling, Yue and Chan-Sew, Derrick and Nakagawa, Rumi and Utpala, Saiteja and Siegel, Zachary S. and Narayanan, Arvind},
	month = jun,
	year = {2026},
	note = {arXiv:2606.26158 [cs.AI]},
	keywords = {Computer Science - Artificial Intelligence},
}