IBM and Notre Dame Launch ‘BenchmarkCards’ to Compare AI Benchmarks

Tech giant IBM (IBM) and the University of Notre Dame have launched a new open-source project to make AI benchmarks easier to understand and compare. While many AI models already come with “model cards” that explain how they were trained and tested, the benchmarks used to judge those models are often poorly documented. To fix this, the researchers created BenchmarkCards, a standardized template and automated tool that clearly explains what each benchmark measures and how it should be used.

The idea behind BenchmarkCards is to give developers a clearer picture of what a benchmark actually tests and where its limits are. Each card includes five sections covering the benchmark’s purpose, data sources, evaluation method, possible risks, and ethical or legal considerations. This makes it easier to compare benchmarks side by side and choose the right one for a specific task. For example, a company trying to reduce harmful content might choose a toxicity-focused benchmark, while a researcher studying fairness would select a bias-focused one.

The researchers hope that this shared format becomes a common language across the AI community. To speed things up, the team also built an automated workflow that can create a benchmark card in about 10 minutes instead of several hours. Indeed, the system pulls information from research papers, converts it into readable text, extracts key details using AI, and then checks those details for accuracy using fact‑verification tools.

Is IBM a Buy, Sell, or Hold?

Turning to Wall Street, analysts have a Moderate Buy consensus rating on IBM stock based on 10 Buys, five Holds, and one Sell assigned in the past three months, as indicated by the graphic below. Furthermore, the average IBM price target of $303.71 per share implies that shares are trading near fair value.

See more IBM analyst ratings