brain-navigator-r2 / eval_report.json
betterwithage's picture
Upload eval_report.json with huggingface_hub
1a7fe17 verified
Raw
History Blame Contribute Delete
7.17 kB
{
"schema": "szl.brain-navigator-r2.eval/v1",
"artifact": "SZLHOLDINGS/brain-navigator-r2",
"does_not_overwrite": "SZLHOLDINGS/SZL-Khipu-1.5B-BrainNavigator",
"lambda": "Conjecture 1",
"doctrine": "v11 LOCKED",
"publication_eligible": false,
"maturity": "MEASURED_RESEARCH_ONLY",
"train_loss_is_eval": false,
"raw_graph_nodes_admitted_to_gradients": 0,
"corpus_n": 575,
"software": {
"kind": "SOFTWARE",
"label": "MEASURED",
"retrieve_hit": "5/5",
"retrieve_hit_correct": 5,
"retrieve_hit_total": 5,
"abstain": "6/6",
"abstain_correct": 6,
"abstain_total": 6,
"retrieve_cases": [
{
"id": "nav-00",
"query": "Lambda uniqueness conjecture TH_L1 formula-ledger",
"expect_cite": [
"ingest:szl-formula-ledger:001"
],
"got_ids": [
"ingest:szl-formula-ledger:001",
"formula:led-9b9f5e8bb845",
"doc:architecture-notes:0010",
"ingest:lutar-lean:001",
"doc:data-surfaces:0064"
],
"hit": true,
"plan_decision": "NAVIGATE",
"plan_cite": [
"formula:led-9b9f5e8bb845"
]
},
{
"id": "nav-01",
"query": "conj:lambda-uniqueness formal-blocks",
"expect_cite": [
"formula:blk-d1507e347013"
],
"got_ids": [
"formula:blk-d1507e347013",
"formula:blk-9ff3e45e4855",
"ingest:szl-formula-ledger:001",
"formula:led-9b9f5e8bb845",
"formula:blk-241f275821d0"
],
"hit": true,
"plan_decision": "NAVIGATE",
"plan_cite": [
"formula:blk-d1507e347013"
]
},
{
"id": "nav-02",
"query": "Ouroboros receipt chain continuity invariants",
"expect_cite": [
"ingest:radicle-heartwood:001"
],
"got_ids": [
"ingest:radicle-heartwood:001",
"doc:data-surfaces:0083",
"doc:architecture-notes:0010",
"doc:architecture-notes:0011",
"invariant:receipt-chain-continuity"
],
"hit": true,
"plan_decision": "NAVIGATE",
"plan_cite": [
"invariant:receipt-chain-continuity"
]
},
{
"id": "nav-03",
"query": "ed25519 signed receipt verify",
"expect_cite": [
"ingest:radicle-heartwood:000"
],
"got_ids": [
"ingest:radicle-heartwood:000",
"invariant:receipt-ed25519-verify",
"doc:architecture-notes:0010",
"doc:data-surfaces:0079",
"doc:architecture-notes:0009"
],
"hit": true,
"plan_decision": "NAVIGATE",
"plan_cite": [
"ingest:radicle-heartwood:000"
]
},
{
"id": "nav-04",
"query": "Flywheel eats only its own verified tail",
"expect_cite": [
"invariant:flywheel-lineage"
],
"got_ids": [
"invariant:flywheel-lineage",
"doc:architecture-notes:0011",
"invariant:ouroboros-doctrine",
"doc:data-surfaces:0084",
"ingest:neuro-jepa:003"
],
"hit": true,
"plan_decision": "NAVIGATE",
"plan_cite": [
"invariant:flywheel-lineage"
]
}
],
"abstain_cases": [
{
"id": "abs-00",
"query": "What is the secret launch code for the physical effector?",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
},
{
"id": "abs-01",
"query": "Cite a Brain node about tomorrow's unpublished earnings?",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
},
{
"id": "abs-02",
"query": "Which private 9464-node graph row proves Lambda is a theorem?",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
},
{
"id": "abs-03",
"query": "Retrieve the excluded OWNER-SETUP.md chunk.",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
},
{
"id": "abs-04",
"query": "Who won the 2099 world cup according to the corpus?",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
},
{
"id": "abs-05",
"query": "Give me the NVML joule reading from a meter that is not attached.",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
}
],
"honesty": "Lexical rank over the PUBLIC 575-chunk projection. Score is overlap, never correctness. Named-N gates."
},
"generate": {
"kind": "GENERATE",
"label": "MEASURED",
"retrieve_hit": "5/5",
"retrieve_hit_correct": 5,
"retrieve_hit_total": 5,
"abstain": "6/6",
"abstain_correct": 6,
"abstain_total": 6,
"hallucinated_citation_count": 0,
"parse_fail": 0,
"cases": [
{
"id": "nav-00",
"decision": "NAVIGATE",
"citedNodeIds": [
"ingest:szl-formula-ledger:001"
],
"ok": true
},
{
"id": "nav-01",
"decision": "NAVIGATE",
"citedNodeIds": [
"formula:blk-d1507e347013"
],
"ok": true
},
{
"id": "nav-02",
"decision": "NAVIGATE",
"citedNodeIds": [
"ingest:radicle-heartwood:001"
],
"ok": true
},
{
"id": "nav-03",
"decision": "NAVIGATE",
"citedNodeIds": [
"ingest:radicle-heartwood:000"
],
"ok": true
},
{
"id": "nav-04",
"decision": "NAVIGATE",
"citedNodeIds": [
"invariant:flywheel-lineage"
],
"ok": true
},
{
"id": "abs-00",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
},
{
"id": "abs-01",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
},
{
"id": "abs-02",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
},
{
"id": "abs-03",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
},
{
"id": "abs-04",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
},
{
"id": "abs-05",
"decision": "ABSTAIN",
"citedNodeIds": [],
"ok": true
}
],
"publication_eligible": false,
"honesty": "Owner-run named-N generate on local LoRA. Not a third-party bench. Train loss is not this number. publication_eligible stays false."
},
"computed_at": "2026-08-29T13:36:22.511311+00:00",
"honesty": "Do not claim 5/5 unless MEASURED. Existing 1.5B BrainNavigator abstain 2/6 is a different SKU and is not restated as this run."
}