Card: Success@5/Recall@1000 columns, mean query length, truncation counts, self-match note only where it applies

README.md

@@ -36,8 +36,8 @@ respectively), see [Encoding](#encoding).
 | `qrels.test.tsv` | text | 8,573 rows | TREC qrels, `qid \t 0 \t docid \t relevance`, no header |
 | `gt_top100.tsv` | text | 293,100 rows | exact MaxSim top-100, see below |
 
-All positional indices (`gt_top100.tsv`, and the row order of every `.npy` file) refer to the order
-of `doc_ids.npy` and `queries_ids.npy`. Reordering either file invalidates `gt_top100.tsv`.
+All positional indices (the `gt_top*.tsv` files, and the row order of every `.npy` file) refer to the
+order of `doc_ids.npy` and `queries_ids.npy`. Reordering either file invalidates the ground truth.
 
 ## Statistics
 
@@ -47,7 +47,7 @@ of `doc_ids.npy` and `queries_ids.npy`. Reordering either file invalidates `gt_t
 | document vectors | 530,764,096 |
 | vectors per document (min / median / mean / max) | 4 / 145 / 218.5 / 8192 |
 | queries | 2,931 |
-| vectors per query (min / median / max) | 9 / 11 / 28 |
+| vectors per query (min / median / mean / max) | 9 / 11 / 11.9 / 28 |
 | queries with at least one qrel | 2,931 |
 | qrels rows | 8,573 |
 | embedding dimension | 128 |
@@ -64,7 +64,7 @@ of `doc_ids.npy` and `queries_ids.npy`. Reordering either file invalidates `gt_t
 | query compute dtype | float32 (model weights loaded at this dtype for the query pass) |
 | query storage dtype | fp32 |
 | normalization | L2, by the model's own `Normalize` module, before the storage cast |
-| document truncation | 8,192 tokens (the checkpoint's `document_length`), before the skiplist; longest document here 8,192 vectors |
+| document truncation | 8,192 tokens (the checkpoint's `document_length`). 79 of 2,428,854 documents (0.0033%) were longer and were cut to it; longest here 8,192 vectors |
 | query truncation | 8,192 tokens (the checkpoint's `query_length`) |
 | document skiplist | none (empty `skiplist_words`): every document token is kept, so doclens is the real token count |
 | document input | `title + "\n\n" + text` when the corpus has a title, else `text`; stripped |
@@ -77,6 +77,7 @@ of `doc_ids.npy` and `queries_ids.npy`. Reordering either file invalidates `gt_t
 ## Ground truth: `gt_top100.tsv`
 
 Exact brute-force MaxSim top-100 per query over the full corpus, from the vectors in this repo.
+
 No header; tab-separated `qidx  docidx  rank  score`:
 
 - `qidx`: 0-based row into `queries_ids.npy` / `queries.npy`
@@ -86,16 +87,14 @@ No header; tab-separated `qidx  docidx  rank  score`:
   dot product`, computed in fp32 with the fp16 document vectors upcast to fp32. Expansion vectors
   are included in the sum. Printed to 6 decimals.
 
-**Self-matches are included.** 1 of 2,931 queries are themselves corpus documents with the same id and retrieve that document (typically at rank 1). This is the raw nearest-neighbour list; BEIR's evaluation drops such pairs (`ignore_identical_ids`), so exclude them before scoring against qrels.
-
 ## Retrieval quality
 
-Sanity check of the vectors, not a leaderboard number: exact MaxSim over the full corpus scored
-against `qrels.test.tsv` with ir_measures, with query-id == doc-id pairs dropped as BEIR does.
+Sanity check of the vectors, not a leaderboard number: `gt_top100.tsv` (exact MaxSim over the full
+corpus) scored against `qrels.test.tsv` with ir_measures.
 
-| nDCG@10 | Recall@100 | MRR@10 | MAP |
-|---|---|---|---|
-| 0.5706 | 0.8369 | 0.6531 | 0.4999 |
+| nDCG@10 | MRR@10 | Success@5 | Recall@100 | Recall@1000 | MAP@100 |
+|---|---|---|---|---|---|
+| 0.5706 | 0.6531 | 0.7919 | 0.8369 | n/a (gt is top-100) | 0.4999 |
 
 ## Loading
 
@@ -152,3 +151,4 @@ Checks run by the exporter on the files exactly as written here:
 |---|---|
 | exported | 2026-09-26 |
 | hardware | Tesla V100S-PCIE-32GB |
+| revised | 2026-09-29: card regenerated; every other file unchanged |