Explore interoperability with EveryEvalEver

[yzeng58]

Source\n\nLeshem Choshen (@LChoshen), an author of Every Eval Ever, commented on Yuchen's BenchPress post asking whether BenchPress data could be combined into EveryEvalEver.\n\n## Paper/project read\n\nEvery Eval Ever is a shared schema, converter/validation pipeline, and community Hugging Face datastore for evaluation result records. It standardizes source metadata, model information, generation/evaluation configuration, metric semantics, aggregate scores, and optional instance-level outputs.\n\nThis appears complementary to BenchPress: BenchPress maintains a curated model-by-benchmark score matrix and prediction layer, while EEE provides a broader interoperable record format and crowdsourced datastore.\n\n## Monthly maintenance action\n\nDuring a future monthly BenchPress maintenance cycle, evaluate whether to:\n\n- export BenchPress public scores to the EEE schema;\n- map BenchPress metadata fields to EEE source/model/config/result blocks;\n- document the relationship between BenchPress and EEE;\n- optionally open an upstream contribution/PR to the EEE datastore.\n