mirror of
https://github.com/garrytan/gbrain.git
synced 2026-07-31 04:07:52 +00:00
Garry Tan
18baca6629
Adds the v1→v2 contract boundary for BrainBench. 6 JSON schemas at
eval/schemas/ pin the shape of every artifact a stack must emit to be
scorable: corpus-manifest, public-probe (PublicQuery with gold stripped),
tool-schema (12 read + 3 dry_run tools, 32K tool-output cap), transcript,
scorecard (N ∈ {1, 5, 10}), evidence-contract (structured judge input).
8 gold file templates at eval/data/gold/ scaffold the sealed qrels,
contradictions, poison items, and citation labels. Empty-but-valid
skeletons; Day 3b fills them with real content once the amara-life-v1
corpus generates.
48 tests validate schema syntax, $schema/$id/title/type headers,
round-trip stability, and cross-schema coherence (new Page types in
manifest enum, tool counts, token cap, N enum).
When v2 ports to Python + Inspect AI + Docker, these schemas are the
boundary. Same fixtures, same tool contracts, zero rework.
16 lines
596 B
JSON
16 lines
596 B
JSON
{
|
|
"version": 1,
|
|
"_comment": "Planted contradictions (10 pairs for amara-life-v1). Each pair states the same fact two different ways in two sources. `canonical` marks which side the scorer trusts; `either-correct` is used sparingly when both are equally valid.",
|
|
"pairs": [
|
|
{
|
|
"_example": "true",
|
|
"id": "c-001",
|
|
"fact": "jordan-park-equity-stake",
|
|
"source_a": {"ref": "emails/em-0012", "claim": "15%"},
|
|
"source_b": {"ref": "meeting/mtg-0003", "claim": "17%"},
|
|
"canonical": "source_b",
|
|
"reason": "meeting notes are later and explicit"
|
|
}
|
|
]
|
|
}
|