Experiment Configuration

Overview

Experiment configuration files provide a YAML-based way to run multiple tasks with shared model settings and per-task overrides. This decouples "how to run" (model, timeouts, agent settings) from "what to test" (task JSON definitions).

Official leaderboard settings

The experiment YAMLs in the all-tasks-k8s example are the official configurations behind the EdgeBench leaderboard numbers.

YAML Format

model:
  api_key: ${OPUS_KEY}
  api_base_url: https://api.anthropic.com
  model: claude-opus-4-6

stagger: 300  # Spread task launches evenly over 300 seconds

defaults:
  agent: claude-code
  timeout: 7200
  eval_interval: 300

tasks:
  ad_placement_optimization:
    eval_interval: 1800
  tinykv:
    extra_env:
      GOPROXY: "https://goproxy.cn"
  gitlet: {}

Fields

model

Configures the LLM used by the agent. All string values support ${ENV_VAR} expansion --- the referenced environment variable must be set at runtime.

Field	Type	Description
api_key	string	API key (supports ${ENV_VAR} expansion)
api_base_url	string	API base URL
model	string	Model name (e.g., claude-opus-4-6, claude-sonnet-4-20250514)

defaults

Default settings applied to all tasks unless overridden per-task.

Field	Type	Default	Description
agent	string	---	Agent name (claude-code, codex, opencode, or a pinned variant such as claude-code-2.1.214)
model	string	---	Model override (overrides model.model for specific tasks if set per-task)
effort	string	---	Reasoning effort (low, medium, high, or max); mapped to the selected Agent when supported
timeout	int	---	Agent timeout in seconds
eval_interval	int	---	Auto-eval interval in seconds
disable_stop_hook	bool	false	Disable the agent stop hook
disable_auto_eval	bool	false	Disable background auto-evaluation
disable_auto_resume	bool	false	Disable auto-resume on abnormal agent exit
internet	bool	---	Enable/disable internet access
extra_env	dict	---	Extra environment variables injected into the agent container
backend	string	---	Container backend (docker, k8s, or e2b)
judge_url	string	---	Judge server URL override
max_submissions	int	---	Maximum number of agent submissions per run
submission_cooldown	int	---	Minimum seconds between agent submissions
work_cpu_limit	int	---	CPU limit for work containers
work_mem_limit	string	---	Memory limit for work containers
judge_cpu_limit	int	---	CPU limit for judge containers
judge_mem_limit	string	---	Memory limit for judge containers

tasks

Per-task overrides. Each key is a task ID that must correspond to a JSON file in the tasks directory. The value accepts the same fields as defaults. Tasks listed here define which tasks are run --- if --task is not specified on the CLI, the experiment's task list is used.

An empty object ({}) means "use all defaults, no overrides."

stagger

A top-level integer field that spreads task launches evenly over the specified number of seconds when running multiple tasks. This prevents a thundering-herd of simultaneous container starts and API calls.

stagger: 300  # Launch tasks at even intervals over 5 minutes

For example, with 6 tasks and stagger: 300, each task starts 50 seconds apart.

Priority Chain

Configuration values are resolved in the following order (highest priority first):

CLI flags  >  per-task overrides  >  experiment defaults  >  env vars / task JSON

For example:

• --timeout 3600 on the CLI always wins
• A per-task timeout: 1800 overrides the experiment-level defaults.timeout
• The experiment defaults.timeout overrides SFORGE_AGENT_TIMEOUT

For extra_env, dictionaries are merged (not replaced). Per-task keys override default keys on conflict:

defaults:
  extra_env:
    FOO: "from-defaults"
    BAR: "shared"

tasks:
  ad_placement_optimization:
    extra_env:
      FOO: "from-task"   # overrides defaults
      BAZ: "task-only"   # added
    # Result: FOO=from-task, BAR=shared, BAZ=task-only

Usage

Run all tasks in an experiment

sforge run --experiment experiments/my_config.yaml --run-id batch-001

When --task is not specified, all tasks listed in the experiment file are run in parallel.

Run a subset of tasks

sforge run --experiment experiments/my_config.yaml --task ad_placement_optimization tinykv

This uses the experiment's model and default settings but only runs the specified tasks.

CLI overrides

sforge run \
  --experiment experiments/my_config.yaml \
  --timeout 3600 \
  --disable-stop-hook \
  --run-id experiment-v2

CLI flags take highest priority and override both per-task and default settings.

Environment Variable Expansion

String values in the model section support ${VAR_NAME} syntax:

model:
  api_key: ${MY_API_KEY}
  api_base_url: ${API_ENDPOINT}

If a referenced variable is not set, SForge exits with an error. This lets you keep secrets out of config files while still having reproducible experiment definitions.

Run Artifacts

When running with an experiment config, SForge saves:

• logs/runs/<run-id>/experiment.yaml --- a copy of the experiment file
• logs/runs/<run-id>/run_config.json --- the fully resolved configuration for all tasks
• logs/runs/<run-id>/<task-id>/run_config.json --- per-task resolved configuration (with API keys redacted)
• logs/runs/<run-id>/summary.json --- final results summary (multi-task runs)

This ensures every run is fully reproducible from its log directory.