diff --git a/.gitattributes b/.gitattributes index 9f45a92b72..48163b46e0 100644 --- a/.gitattributes +++ b/.gitattributes @@ -1 +1 @@ -evals/registry/data/**/*.jsonl filter=lfs diff=lfs merge=lfs -text +evals/registry/data/**/*.jsonl filter=lfs diff=lfs merge=lfs -text \ No newline at end of file diff --git a/evals/registry/data/population_span_extraction/samples.jsonl b/evals/registry/data/population_span_extraction/samples.jsonl new file mode 100644 index 0000000000..7c58463994 --- /dev/null +++ b/evals/registry/data/population_span_extraction/samples.jsonl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d75d73da796eacae15d03b59da8578f496ca3b6450d9a77cb1d28670e558e628 +size 2323327 diff --git a/evals/registry/evals/population_span_extraction.yaml b/evals/registry/evals/population_span_extraction.yaml new file mode 100644 index 0000000000..a6a3b62e14 --- /dev/null +++ b/evals/registry/evals/population_span_extraction.yaml @@ -0,0 +1,13 @@ +# Register the eval by adding a file to evals/registry/evals/.yaml +population_span_extraction: + id: population_span_extraction.dev.v0 + metrics: [accuracy] + description: The model is shown abstracts of clinical drug trials and tasked with extracting the text spans that specify the population demographic of the shown abstract. The population demographic can be but is not necessarily specified in multiple seperate spans. + +population_span_extraction.dev.v0: + # Assuming that the response can contain "significant variation" --> modelbased classify + class: evals.elsuite.modelgraded.classify:ModelBasedClassify + args: + samples_jsonl: population_span_extraction/samples.jsonl + eval_type: cot_classify + modelgraded_spec: fact \ No newline at end of file