Open IETF draft
Can your agent implement an IETF draft?
Start with our prompt.
SER is a published Internet-Draft, not an approved IETF standard.
We are developing an open format for signed evidence of AI evaluations. We are now looking for an independently developed implementation from outside our project.
Benchmark scores travel today as screenshots, tables or blog posts. A valid SER receipt binds the recorded payload to a verification key through a signature. Who controls that key, whether they are authorised to make the claim and whether the result is true must be checked separately.
An implementation from outside our project helps us identify which rules are clear from the text alone and where interpretations differ.
Your results help us find ambiguities and improve the draft. After reviewing the results together, we aim to document your implementation with your permission.
Your code. Your reading. A contribution that counts.
Build your own verifier with your coding agent, from the published draft alone.
You help find out which rules are unambiguous and where different readings arise.
Send us the report and test results. We will send a new challenge set for your batch command. After reviewing the results together, we aim to document your implementation in the next revision with your permission.
Run the new challenge set first with the version recorded in your report and send back the results. We compare the answers and do not run any code you send. Document any changes to the verifier separately.
We aim to include jointly reviewed reports in the next revision after IETF 127.
Sponsor