Skip to content
The Chat Archive
gpt-4.1-mini-2025-04-14· OpenAI LMArena Human Preference 140K

give me a hard verifiable prompt to compare the abilities of llm

  1. User

    give me a hard verifiable prompt to compare the abilities of llm

  2. Assistant

    Certainly! Here's a hard, verifiable prompt designed to rigorously test and compare the reasoning, knowledge, and synthesis abilities of large language models (LLMs). It requires multi-step reasoning, factual knowledge, and clear explanation, making it a strong benchmark for comparison:


    Prompt:

    "Explain the key differences between classical Newtonian mechanics and Einstein's theory of General Relativity in terms of their treatment of gravity. Then, provide a detailed example of a physical phenomenon that Newtonian mechanics cannot accurately describe but General Relativity can. Finally, calculate the approximate amount of gravitational time dilation experienced by a clock located 10 km above the surface of the Earth compared to a clock on the surface, using the Schwarzschild metric approximation. Show all steps and assumptions clearly."


    Why this prompt is hard and verifiable:

    1. Conceptual understanding: Requires clear explanation of two major physical theories and their conceptual differences.
    2. Application: Needs an example where Newtonian mechanics fails but GR succeeds (e.g., GPS satellite time dilation, Mercury’s perihelion precession).
    3. Mathematical reasoning: Involves applying the Schwarzschild metric formula for gravitational time dilation, requiring correct use of constants and approximations.
    4. Verification: The explanation and calculations can be checked against physics textbooks and reliable sources.

    If you want, I can also provide a model answer to this prompt for benchmarking purposes. Would you like that?

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.