MicroEvals

Run your prompts across multiple models to compare their performance.

History

Models

Prompt

Examples

Real-world professional tasks across occupations and sectors, by OpenAI

Loading questions…

Public evaluations

Showing 181-200 of 5354

👍1

构造一个最大值小于等于m的序列，满足任意2数的gcd不重复，求最长长度

👍1

Web Dev w/ Non-Detailed Prompts

This microeval will evaluate how well leading coding models can develop different types of web apps based on prompts that aren't ultra-specific, and just ask for the overall concept. This is to test how much LLMs have evolved in terms of design skills and common knowledge coding choices. There is one detailed prompt to see if it enhances the quality.

👍1

Trading Instruction Fidelity (TIF-5)

Tests whether a model follows explicit, verifiable instructions in trading and investing contexts without access to real-time market data. Each prompt carries deterministic pass/fail constraints — exact counts, banned words, ordering, output templates, and math on provided data — so grading measures instruction compliance, not market knowledge or data access.

Generator

Strawrbrerrry [sic] eval

Reasoning should include the ability to generalize to unfamiliar words instead of memorizing answers. Let's see if models can detect the number of 'r's in the word "strawrbrerrry."