Source record

Published by OpenAI News (T1 source). The original is at https://openai.com/index/terminal-x.

Pipeline notes

The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.

Terminal X's Eval Agent uses GPT-6.1 Sol to grade outputs against a 54-criterion rubric. The system reportedly matches senior investor judgment 88.6% of the time and catches 25% more deviations from a fund's mandate than prior approaches.

Concrete accuracy figures for rubric-based agent evaluation in a finance workflow, with a stated delta over earlier methods. Useful baseline for anyone designing agent eval loops.

Cited by

No citations on record.