Skip to content
This repository was archived by the owner on Mar 6, 2026. It is now read-only.

Releases: lorehq/delegation-study

Study 1: Session Data (280 sessions)

01 Mar 03:08

Choose a tag to compare

Raw session data for all 280 sessions across 7 conditions and 6 tasks.

Contents:

  • session.jsonl: Full JSONL transcripts of each Claude Code session
  • output.json: Structured output with cost, duration, model usage
  • task.json: Task metadata per run
  • battery-results.json: Per-batch audit summaries

Archive structure: .runs/battery/////

To use: extract into the repository root and run analysis scripts:
tar xzf session-data-v1.tar.gz
node harness/stats.js
node harness/score-correctness.js