Show HN: Benchmarklist: track AI benchmarks (2.4k+), models, and capabilities (benchmarklist.com)

A new platform called Benchmarklist has been launched to track over 2,400 AI benchmarks, models, and capabilities, enhancing accessibility for researchers and developers in the AI/ML community. This significant resource consolidates disparate benchmark data from various domains, streamlining the benchmarking process and fostering a collaborative environment for performance analysis across numerous AI models and tasks. Among the notable benchmarks included are RoboDojo—a unified evaluation framework for generalist robot manipulation that blends simulation and real-world tasks—and SALT, which focuses on assessing the accuracy and reliability of long-form outputs from large language models (LLMs) with precision. RoboDojo supports 60 tasks across five dimensions, while SALT provides a deterministic framework for evaluating LLMs, making it easier to identify encoding errors and improve output quality. These benchmarks address a critical need for systematic evaluation metrics that can confirm and reproduce performance across various AI systems while enabling clearer comparisons and improvements in model design and deployment strategies.