Life After Benchmark Saturation: A Case Study of CORE-Bench
Quick Answer
CORE-Bench Hard reveals that after accuracy saturation, evaluating agent performance on dimensions like efficiency and reliability provides deeper insights.
Quick Take
The introduction of CORE-Bench v1.1 and CORE-Bench OOD enhances measurement capabilities, showing significant performance uplift from human-agent collaboration, with speed improvements around twofold.
Key Points
- CORE-Bench Hard identifies construct validity threats in agent performance evaluation.
- CORE-Bench v1.1 and CORE-Bench OOD improve benchmarking for efficiency and reliability.
- Human-agent collaboration yields a statistically significant speedup, approximately twofold.
- Accuracy saturation does not diminish the relevance of CORE-Bench v1.1 for performance metrics.
- The study advocates for a broader evaluation framework beyond accuracy-centric approaches.
Paper Resources
📖 Reader Mode
~2 min readAuthors: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.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.26158 [cs.AI] |
| (or arXiv:2606.26158v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2606.26158 arXiv-issued DOI via DataCite |
Submission history
From: Nitya Nadgir [view email]
[v1]
Tue, 23 Jun 2026 22:30:44 UTC (2,036 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising
HOBA (Hierarchical On-policy Bidding Agents) is a novel hierarchical reinforcement learning framework that enhances online advertising bidding systems by improving adaptability and reducing hyperparameter tuning costs. It utilizes a for hyperparameter inference, a SARSA agent for expert model selection, and a dynamic expert pool for bid execution, achieving a +3.6% increase in target cost during large-scale deployment and outperforming state-of-the-art baselines on AuctionNet.