Skip to content

Self-Improvement Agent Harness: A Deterministic SIA Exemplar (v0.1.1)

Latest

Choose a tag to compare

@docxology docxology released this 14 Jun 19:58
· 1 commit to main since this release

Release v0.1.1 for templates/template_sia.

Publication

Abstract

Abstract

This exemplar documents template_sia, a deterministic implementation of the Self-Improvement Agent (SIA) harness contract described in . The default pipeline replays fixture-backed generations for the mini_classify task; opt-in live mode runs bounded target subprocesses and optional Ollama-backed meta/feedback steps.

Run snapshot. Task mini_classify, run 1, 3 generation(s), live=false. Final accuracy=0.8333 over 6 held-out samples. Values are injected by scripts/z_generate_manuscript_variables.py after analysis.

Keywords: self-improvement agents, benchmark harness, reproducible evaluation, agent loops