FrontierScience
A scientific reasoning evaluation suite released by OpenAI that covers frontier problems in disciplines such as physics, chemistry, and biology. It emphasizes multi-step reasoning and specialized knowledge, and is characterized by high difficulty and strong discriminative power.
Timeline (last 90 days)
On the FrontierScience benchmark developed by OpenAI, using data generated by FormulaForest raised the accuracy of the Qwen-35B-A3B model from 61% to 72%.
The FrontierScience benchmark is a scientific reasoning evaluation set released by OpenAI, covering frontier problems in disciplines such as physics, chemistry and biology, and emphasizing difficult problems that require multi-step reasoning and specialized knowledge.
All Facts (2)
On the FrontierScience benchmark developed by OpenAI, using data generated by FormulaForest raised the accuracy of the Qwen-35B-A3B model from 61% to 72%.
50%UnverifiedThe FrontierScience benchmark is a scientific reasoning evaluation set released by OpenAI, covering frontier problems in disciplines such as physics, chemistry and biology, and emphasizing difficult problems that require multi-step reasoning and specialized knowledge.
50%