Reference¶

Look up the mechanics of Agentic Arena: run modes, comparison controls, evaluation definitions, commands, and repository structure.

Most-used references¶

Need	Reference
Understand what a result can prove	Run modes
Interpret profile evidence and freshness	Profile evidence labels
Reproduce or audit a comparison	Methodology
Check controlled inputs	Fairness controls
Run with a subscription-backed model	Codex bridge
Inspect adapter dependencies	Dependencies
Understand a scorecard artifact	Scorecard example
See planned project work	Next phases

Command map¶

python -m arena validate
python -m arena run --arena tool_use --framework vanilla --mode mock
python -m arena scorecard --arena tool_use
python -m arena summary --print
python -m arena chart --arena tool_use

Repository map¶

Path	Responsibility
`arena/`	harness, configuration, runner, scorer, and reports
`arena/llm/`	OpenAI-compatible client, mock server, and Codex bridge
`arena/tools/`	shared tools and fixed local corpus
`arenas/<id>/`	workload specification, dataset, and mock script
`frameworks/<name>/`	one framework adapter and its dependencies
`results/`	committed native live scorecards
`runs/`	local and CI artifacts that are not published as results