wildtrace / eval /config.json
CinderD's picture
Add Gemini-native judge compatibility to eval harness
9c7ab6a verified
Raw
History Blame Contribute Delete
1.88 kB
{
"_comment": "WildTrace eval/judge config. Point base_url at ANY OpenAI-compatible /chat/completions endpoint and set model. api_key_env names the env var holding the bearer token. See methodology/EVAL_PROTOCOL.md for the exact protocol and the per-model context caps used in the paper.",
"base_url": "https://api.openai.com/v1/chat/completions",
"api_key_env": "API_KEY",
"model": "your-model-id",
"max_tokens": 32768,
"send_temperature": true,
"temperature": 0.1,
"timeout_s": 900,
"_caps_note": "Per-model context caps in CHARACTERS (en docs / cjk docs). A doc longer than the cap is out_of_context_scope (scored 0, not sent). If your model id is not listed here or in run_eval.py's DEFAULT_CAPS, set cap_key to 'default' (2.85M en / 850K cjk) or add your model's REAL native window. Do NOT inherit an older version's cap.",
"cap_key": "default",
"context_caps": {
"default": {"en": 2850000, "cjk": 850000}
},
"_judges_note": "3 non-contestant judges, averaged. The paper used Claude-Sonnet-4.6, Qwen3.5, and Gemini-2.5-Flash. Chat endpoints are the default; for gateways exposing Gemini through native contents/candidates payloads, set api_type to gemini_native on that judge. api_key_env defaults to the top-level one if omitted. List one judge to grade with a single judge (the paper headline is the 3-judge average).",
"judges": [
{"name": "sonnet", "base_url": "https://api.openai.com/v1/chat/completions", "model": "claude-sonnet-4-6", "api_key_env": "API_KEY", "max_tokens": 16384},
{"name": "qwen3.5", "base_url": "https://api.openai.com/v1/chat/completions", "model": "qwen3.5-plus", "api_key_env": "API_KEY", "max_tokens": 16384},
{"name": "gemini-flash", "base_url": "https://api.openai.com/v1/chat/completions", "model": "gemini-2.5-flash", "api_key_env": "API_KEY", "max_tokens": 16384}
]
}