Search Results for "mix"

...It’s built to act like a research agent: synthesizing, reasoning, retrieving information via the web and documents, and backing its outputs with evidence. The model is about 30.5 billion parameters in size, though at any given token only ~3.3B parameters are active. It uses a mix of synthetic data generation, fine-tuning and reinforcement learning; supports benchmarks like web search, document understanding, question answering, “agentic” tasks; provides inference tools, evaluation scripts, and “web agent” style interfaces. The aim is to enable more autonomous, agentic models that can perform sustained knowledge gathering, reasoning, and synthesis across multiple modalities (web, files, etc.).

BIG-bench (Beyond the Imitation Game Benchmark) is a large, collaborative benchmark suite designed to probe the capabilities and limitations of large language models across hundreds of diverse tasks. Rather than focusing on a single metric or domain, it aggregates many hand-authored tasks that test reasoning, commonsense, math, linguistics, ethics, and creativity. Tasks are intentionally heterogeneous: some are multiple-choice with exact scoring, others are free-form generation judged by...