Vector drift study DB snapshot

Database snapshot for the study "Vector Retrieval Drift under Corpus Growth" (EnterpriseRAG-Bench, 5k→100k gold-pinned ladder).

Asset	Contents
erb_tables.dump.gz	pg_dump (custom format): OpenAI text-embedding-3-small document embeddings (document_chunks), runs, per-query results
erb_titan_tables.dump.gz	pg_dump (custom format): Amazon Titan Text Embeddings V2 arm (document_chunks_titan, experiment_runs_titan, vector_drift_results_titan)
erb_query_embed_cache.json.gz	Cached primary-200 query embeddings (OpenAI)
erb_titan_query_embed_cache.json.gz	Cached primary-200 query embeddings (Titan V2)

Restore (see docs/REPLICATE.md):

./scripts/erb/restore_from_backup.sh data/backups/erb_tables.dump.gz
./scripts/erb/restore_from_backup.sh data/backups/erb_titan_tables.dump.gz   # optional Titan arm

SHA-256:

• erb_tables.dump.gz 5bf3a324f803a6c478b54bb62214357507c815bb747550863860b32b7ea5c02e
• erb_titan_tables.dump.gz f3e25131ec88a43fece1536f4d9e692d4fc83d6ea209d261acfe7f63f18992da
• erb_query_embed_cache.json.gz e3878269ce2f6b077f4699022474b88b059b5c5d240f20ab0ac249c66d0156a4
• erb_titan_query_embed_cache.json.gz 057a42a5794569e5a99eff5c44c7ffe2abb152eb02cc1701d237295543dcb12c

Row counts: document_chunks 103,033 (99,998 with scale_rank < 100k); document_chunks_titan 99,998.