- Schema: dense-rar-v3
- Run ID: dense-rar-2026-04-24T0801
- Mode: formal (live providers explicitly allowed)
- Git: 3c274d1 (clean)
- Vault root: <REMOTE_RESEARCH_VAULT>
- Manifest: 200 entries (18 real + 182 adversarial junk)
- Manifest hash: bd9afb901dadd199
- Queries hash: bcce230b845b7226
- Query suite: wave3-adversarial-retrieval (1.0.0, adversarial)
- Query categories: near_neighbor_memory, lexical_trap, cross_lingual, negative_exclusion, methodology_philosophy, engineering_specific
- Queries: 24
- Runs per (pipeline × query): 1
- Stage 1: BGE-M3-Q4_K_M-8080 via <NON_LOCAL_PROVIDER_ENDPOINT> (model: bge-m3)
- Embedding timeout: 180000ms
- Corpus batch size: 50
- Stage-1 candidate pool: 20
- Final top-K: 5
- Stage 2 judge: deepseek-v4-flash via https://api.deepseek.com (thinking: disabled)
- Pipelines: dense, dense-rar, dense-adaptive
- Total calls: 72
| Pipeline | Status | Valid | Errors | Top-1 | Top-5 | p50 latency | p90 latency | Avg latency | Judge exposure |
|---|---|---|---|---|---|---|---|---|---|
| dense | ok | 24/24 | 0 | 16/24 (66.7%) | 18/24 (75.0%) | 520ms | 552ms | 5666ms | 0 |
| dense-rar | ok | 24/24 | 0 | 23/24 (95.8%) | 23/24 (95.8%) | 1533ms | 1761ms | 1600ms | 20 |
| dense-adaptive | ok | 24/24 | 0 | 23/24 (95.8%) | 23/24 (95.8%) | 1528ms | 1713ms | 1580ms | 20 on 100%, 0 on 0% |
| Pipeline | Category | Valid | Top-1 | Top-5 | Candidate hit | Candidate miss | Synthetic miss | Real miss | Empty miss |
|---|---|---|---|---|---|---|---|---|---|
| dense | cross_lingual | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense | engineering_specific | 4/4 | 1/4 | 1/4 | 4/4 | 0/4 | 3 | 0 | 0 |
| dense | lexical_trap | 4/4 | 3/4 | 4/4 | 4/4 | 0/4 | 1 | 0 | 0 |
| dense | methodology_philosophy | 4/4 | 3/4 | 4/4 | 4/4 | 0/4 | 1 | 0 | 0 |
| dense | near_neighbor_memory | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense | negative_exclusion | 4/4 | 1/4 | 1/4 | 3/4 | 1/4 | 3 | 0 | 0 |
| dense-rar | cross_lingual | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense-rar | engineering_specific | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense-rar | lexical_trap | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense-rar | methodology_philosophy | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense-rar | near_neighbor_memory | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense-rar | negative_exclusion | 4/4 | 3/4 | 3/4 | 3/4 | 1/4 | 1 | 0 | 0 |
| dense-adaptive | cross_lingual | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense-adaptive | engineering_specific | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense-adaptive | lexical_trap | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense-adaptive | methodology_philosophy | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense-adaptive | near_neighbor_memory | 4/4 | 4/4 | 4/4 | 4/4 | 0/4 | 0 | 0 | 0 |
| dense-adaptive | negative_exclusion | 4/4 | 3/4 | 3/4 | 3/4 | 1/4 | 0 | 0 | 1 |
- Result status: valid-execution
- Formal eligible: yes
- Skip rate: 0/24 (0.0%)
- Top-1 on skipped queries: 0/0 (0%)
- Top-1 on invoked queries: 23/24 (95.8%)
- Gate threshold: stage1_score[0] / stage1_score[1] ≥ 1.5