Skip to content

Inconsistency in benchmark? #3

Description

@rodrigo-castellon

This might be a misunderstanding on my part, please correct me if I'm wrong.

I'm seeing here in the standard GSM8K dataset (http://platinum-bench.csail.mit.edu/inspect?model=gemini-2.5-pro-exp-03-25&dataset=gsm8k) that many of the LLMs are getting this problem wrong:

Image

It looks to me like the problem's solution is incorrect. 80 flagstones would be 80 * 75 = 6,000 pounds, which would require 3 trucks (each carries 2,000 pounds).

However, in GSM8K-Platinum, I also see this problem come up with the same listed solution:

Image

That being said, some of the models (like Gemini 2.5 Pro) are not listed (presumably due to having 0 errors?)

Thanks for your time.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions