Physical AI models are evaluated on suites of benchmarks that differ across model reports, leaving the model-by-benchmark matrix sparse and the relationship between benchmarks unmeasured. We construct...
The paper introduces Balance of Benchmarks (BoB), a framework that improves task-conditioned model comparison by weighting benchmark evidence based on semantic density, equating scores across varying difficulty levels, and pooling task-relevant residuals. BoB retains all eligible benchmark data while adjusting its influence, outperforming uniform averaging on the WildScores dataset with higher Spearman correlation, lower MAE, and better shortlist hit rates. The method also reduces ranking instability when benchmarks are repeated or paraphrased, and lowers retrospective regret in model selection.
arXiv:2605. 17273v3 Announce Type: replace-cross Abstract: State-of-the-Art (SOTA) claims pervade Artificial Intelligence (AI) and Machine Learning (ML) research.
arXiv:2606. 24020v1 Announce Type: new Abstract: A modern model release reports scores on 40+ benchmarks and the same evaluations were run many more times before it: to track training progress, compare design choices, and select the checkpoint for the release.
arXiv:2607. 02032v1 Announce Type: new Abstract: Evaluating LLM agents on benchmarks like SWE-Bench and GAIA can be expensive, time-consuming, and requires complex infrastructure.
arXiv:2607. 25891v1 Announce Type: new Abstract: Evaluating AI agents in interactive environments is hindered by fragmented tasks, scaffolds, verifiers, and scoring rules.
arXiv:2607. 15190v1 Announce Type: new Abstract: AI benchmarks increasingly leverage item-level statistical models, particularly item response theory (IRT), to estimate model capabilities, rank systems, select informative examples, and diagnose benchmark quality.
arXiv:2608.30044v1 Announce Type: new Abstract: Language models are commonly compared by averaging scores across a benchmark list with equal weight. Such lists grow through publication outside an exp...
The paper examines how the choice of anchor model in LLM-as-a-judge evaluations affects reliability. By testing 22 anchors on the Arena-Hard-v2.0 dataset, it shows that extreme anchors (best or worst performers) are poor choices, reducing correlation with human rankings. The study quantifies the anchor effect size, compares it to judge model selection, and offers guidelines and power‑analysis recommendations for more reliable benchmark design.
The article argues that for frontier language models, precision—how consistently outputs cluster around the target—should be the key metric rather than capability, which measures average performance. It proposes a simple, non‑circular method to quantify precision by repeatedly scoring deterministic tasks and computing outcome consistency, and demonstrates how this metric can guide decisions about model improvements. The study shows that precision can reveal whether failures are due to systemic misalignment or random noise, and that real‑world measurement is more valuable than rule‑based benchmarks.
arXiv:2608.21601v1 Announce Type: new Abstract: Benchmarks for scientific artificial intelligence are mostly written to be scored: multiple-choice questions, curated agent tasks with reference soluti...
The paper introduces a low‑rank framework for ranking large language models (LLMs) on task‑specific benchmarks using sparse pairwise comparisons. By modeling the task‑by‑model ability matrix as low rank, the method shares information across related tasks while preserving task‑specific differences, and it provides uncertainty‑aware ranking through debiased estimators and simultaneous confidence sets. Experiments on synthetic data and the Chatbot Arena benchmark demonstrate improved sample efficiency and tighter, better‑calibrated ranking certificates, especially in the sparse comparison regime typical of real LLM evaluations.