Loading...
Loading...
Contributors:
Dataset URI:
10.5281/zenodo.23334128# Proposer-pool diversity experiment (E2-A): ledger, frozen protocol, and analysis Artifact for the proposer-pool diversity experiment, domain A (job-shop scheduling), of the verifier-bound automated research program. The experiment tests whether pool-composition diversity -- manipulated at a FIXED verifier -- changes discovery quality, the first-order variable identified by E1. ## Contents - `PREREGISTRATION.md` -- frozen protocol (E2-A domain A and E2-D domain D; E2-D ledger is released separately upon completion). - `e2_a.jsonl` -- per-pair ledger, one JSONL line per (arm, dose, pair): seed, proposals, promotions, holdout values. Append-only. - `analyze_e2.py` -- frozen analysis script (H2/H3 gates and paired contrasts). All reported numbers are script-generated. - `analyze_e2_results.json` -- analysis output. ## Provenance - Doses D0/D1/D2 manipulate pool diversity at a fixed S0 verifier; random arm (seed 9000+/9500+ space, disjoint from E1) is the diversity-free control. - Holdout instances evaluated exactly once per promoted rule (HoldoutGate, one-shot semantics). ## License Data and protocol CC-BY-4.0; analysis code MIT.
Keyword(s):
LLM agents; scientific discovery; pool diversity; pre-registration; job-shop scheduling.
MeSH Term(s):
Grants and Funding:
No Grants or Funding Data Available
Associated Publication(s):
No Associated Publication Data Available
Supplementary Data:
No Supplementary Data Available
Related Dataset(s):
No Related Dataset Data Available
The National Library of Medicine (NLM) provides information about biomedical datasets. Inclusion on an NLM Web page does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: NLM Web Policies (nih.gov)