# sunny-glow/Auto-BenchMax

Repository: https://github.com/sunny-glow/Auto-BenchMax
Canonical: https://ross.abutalabs.com/products/auto-benchmax
Language: Python
License Family: other
Last push: 2026-07-24T03:52:24+00:00

## Health v2 (maintenance only)
Score: 55/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 94, release rhythm 35, longevity 2
- inputs: {"age_days": 41, "days_push": 40, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, young, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1317, forks 27 (observed 2026-08-28T04:04:20.595856+00:00)

## What it is
Auto-BenchMax is a Python pipeline that automatically synthesizes benchmark-targeted training data for LLMs, claiming to more than double a base model's score on agentic benchmarks like MCP-Atlas and Tau2-bench. It ships a data-construction pipeline, training scripts, an evaluation environment, and a skill that lets agent tools like Claude Code drive synthesis for custom tool-use benchmarks.

## Use cases
- synthesize training data for an agentic benchmark
- improve my model's score on MCP-Atlas
- generate fine-tuning data targeted at a tool-use benchmark
- reproduce benchmark score improvements with one click
- create LLM-judged or rule-based benchmark training sets
- fine-tune a model to double its benchmark baseline

## When to choose
- you need to boost a model's score on a specific tool-use or agentic benchmark
- you want a reproducible data-synthesis plus training pipeline
- you use a skill-capable agent like Claude Code and want one-sentence automation

## When to avoid
- you need genuinely general capability gains rather than benchmark-targeted optimization
- you lack the compute to fine-tune models
- you need a license-cleared project for commercial use, since no license is specified

## Facets
- artifact type: library
- maturity: active
- function: llm-training, data-generation, agent-framework, machine-learning
- domain: large-language-models, machine-learning, artificial-intelligence, developer-tools
- platform: python, cli
- tags: benchmark-optimization, synthetic-data, fine-tuning, tool-use, skill-based-pipeline, llm-judge, training-scripts, ai-agents, linux

## Member repositories
- sunny-glow/Auto-BenchMax (main) score 55

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:20.595856+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:48:14.970053+00:00, confidence not recorded.
  - readme: https://github.com/sunny-glow/Auto-BenchMax (fetched 2026-08-28T04:04:20.595856+00:00, sha a6fbf9ceb2e0)
- Data as of 2026-08-30T08:39:29.467469+00:00.