Hugging Face's logo

Datasets:
hakari-bench
/
results

Dataset card Files
xet
 Community
47

Add results for sbintuitions/sarashina-embedding-v2-1b

#21
by hotchpotch - opened
base: refs/heads/main
←
from: refs/pr/21
Discussion Files changed
+1614
-0
Add results for sbintuitions/sarashina-embedding-v2-1bb3d8533a
hotchpotch
 HAKARI-Bench org

Add HAKARI-Bench results for sbintuitions/sarashina-embedding-v2-1b

Summary

Field Value
Model sbintuitions/sarashina-embedding-v2-1b
Result directory sbintuitions__sarashina-embedding-v2-1b
Target path hakari-results/sbintuitions__sarashina-embedding-v2-1b
Result files 538 total, 538 .json.xz
Evaluation method dense
Overall nDCG@10 0.2499
Overall score units 369 grouped units from 525 raw task results

DuckDB Nano-set Comparison

Computed from DuckDB task_results with the same Overall grouping as this PR body. Quantized and rescore variants are excluded; truncate variants are considered, and each model column uses that model's best Overall variant.

Overall component sbintuitions/sarashina-embedding-v2-1b Qwen/Qwen3-Embedding-0.6B (1024 dims) jinaai/jina-embeddings-v5-text-small (1024 dims) BAAI/bge-m3 (1024 dims) intfloat/multilingual-e5-small (384 dims) bm25
Overall 0.2499 0.5979 0.6323 0.5859 0.5190 0.4832
NanoMMTEB-v2 0.3148 0.5581 0.5590 0.4846 0.4455 0.4550
NanoRTEB 0.2999 0.6713 0.7005 0.5365 0.4711 0.3553
MNanoBEIR 0.2066 0.5509 0.6077 0.5575 0.5117 0.4646
NanoBIRCO 0.1562 0.3070 0.3526 0.2617 0.1613 0.2693
NanoMLDR 0.1668 0.6239 0.5384 0.6621 0.3920 0.7396
NanoLongEmbed 0.3300 0.7232 0.6680 0.6527 0.5014 0.8217
NanoDAPFAM 0.1607 0.3018 0.3179 0.2406 0.2380 0.2400
NanoCoIR 0.5590 0.8601 0.8777 0.6924 0.6915 0.5436
NanoIFIR 0.0876 0.3364 0.3893 0.2391 0.2152 0.2761
NanoLaw 0.3005 0.6075 0.6370 0.5597 0.4790 0.6854
NanoMedical 0.2041 0.5694 0.5803 0.5371 0.5055 0.4145
NanoRARb 0.1022 0.2689 0.2889 0.2343 0.2240 0.1359
NanoBRIGHT 0.2252 0.3885 0.4284 0.2941 0.1758 0.2790
NanoCodeRAG 0.4772 0.8712 0.9139 0.7155 0.7464 0.5823
NanoChemTEB 0.4113 0.8035 0.7980 0.7777 0.8081 0.7012
NanoR2MED 0.0978 0.3180 0.3630 0.2088 0.1099 0.2094
NanoBuiltBench 0.2873 0.5129 0.5277 0.4248 0.4291 0.3958
NanoCMTEB 0.5133 0.7982 0.8052 0.7591 0.6999 0.6003
NanoIndicQA 0.1034 0.6413 0.7056 0.7586 0.7009 0.5653
NanoMuPLeR 0.1260 0.7122 0.8388 0.8912 0.7837 0.7994
NanoMTEB-v2 0.3522 0.6372 0.6450 0.5726 0.5348 0.5028
NanoMTEB-Dutch 0.1905 0.5686 0.6213 0.5863 0.5287 0.4673
NanoMTEB-French 0.2254 0.5771 0.6377 0.5527 0.4702 0.4261
NanoMTEB-German 0.2001 0.6298 0.6536 0.6189 0.5711 0.5522
NanoJMTEB-v2 0.7307 0.7732 0.8008 0.7906 0.7165 0.7465
NanoMTEB-Korean 0.5465 0.7792 0.8246 0.8183 0.7668 0.6743
NanoFaMTEB-v2 0.1678 0.6338 0.6882 0.6652 0.6135 0.5651
NanoMTEB-Polish 0.1162 0.4738 0.5316 0.4999 0.4365 0.3424
NanoRuMTEB 0.3912 0.8622 0.9121 0.9169 0.8643 0.7089
NanoMTEB-Scandinavian 0.2463 0.6981 0.7596 0.7740 0.7029 0.6091
NanoMTEB-Spanish 0.1993 0.5662 0.6292 0.5624 0.4848 0.3679
NanoMTEB-Thai 0.2001 0.7455 0.7670 0.7672 0.7107 0.5216
NanoVNMTEB 0.1917 0.5717 0.6066 0.5616 0.5197 0.4571
NanoMTEB-Misc 0.3065 0.7629 0.8011 0.7766 0.6423 0.4939
NanoMIRACL 0.2951 0.7879 0.8351 0.8475 0.7871 0.5715

Overall nDCG@10

Overall component nDCG@10 Score units Raw task results
NanoMMTEB-v2 0.3148 18 18
NanoRTEB 0.2999 14 14
MNanoBEIR 0.2066 13 169
NanoBIRCO 0.1562 5 5
NanoMLDR 0.1668 13 13
NanoLongEmbed 0.3300 6 6
NanoDAPFAM 0.1607 12 12
NanoCoIR 0.5590 10 10
NanoIFIR 0.0876 4 4
NanoLaw 0.3005 4 4
NanoMedical 0.2041 7 7
NanoRARb 0.1022 14 14
NanoBRIGHT 0.2252 20 20
NanoCodeRAG 0.4772 4 4
NanoChemTEB 0.4113 3 3
NanoR2MED 0.0978 8 8
NanoBuiltBench 0.2873 2 2
NanoCMTEB 0.5133 8 8
NanoIndicQA 0.1034 11 11
NanoMuPLeR 0.1260 14 14
NanoMTEB-v2 0.3522 10 10
NanoMTEB-Dutch 0.1905 27 27
NanoMTEB-French 0.2254 8 8
NanoMTEB-German 0.2001 5 5
NanoJMTEB-v2 0.7307 11 11
NanoMTEB-Korean 0.5465 5 5
NanoFaMTEB-v2 0.1678 17 17
NanoMTEB-Polish 0.1162 14 14
NanoRuMTEB 0.3912 3 3
NanoMTEB-Scandinavian 0.2463 7 7
NanoMTEB-Spanish 0.1993 7 7
NanoMTEB-Thai 0.2001 9 9
NanoVNMTEB 0.1917 26 26
NanoMTEB-Misc 0.3065 12 12
NanoMIRACL 0.2951 18 18

Reproducibility

Field Value
Model source sbintuitions/sarashina-embedding-v2-1b
Model revision 1f3408afaa7b617e3445d891310a9c26dd0c68a5
Dataset revision(s) 01736efbaa96f020c2a4d996efdacc18071e2fcb, 017849a95097eea984680cbab35972f8d3812376, 0f3a6f43b8a26a9b8c8d5f31b09bd60dc4cd572d, 1726763179e1e114ad9ffcdc7262923471e8ecc8, 175ff423246cdbca9c3a992c4d68d312701b3f2a, ... (47 total)
Evaluated at UTC 2026-07-15T01:07:19.953369+00:00 to 2026-07-15T13:59:37.657356+00:00
Generated at UTC 2026-07-15T01:07:20.219590+00:00 to 2026-07-15T13:59:37.657374+00:00
dtype bf16
device cuda:0
batch size 16, 32, 8
attention implementation sdpa
trust remote code False
max sequence length 8192
candidate ranking reranking_hybrid
rerank top-k not recorded
query prompt name not recorded
document prompt name not recorded
Python 3.12.12 (main, Dec 9 2025, 19:02:36) [Clang 21.1.4 ]
Platform Linux-6.8.0-107-generic-x86_64-with-glibc2.39
torch 2.9.0
transformers 5.12.1
sentence-transformers 5.4.1
datasets 4.8.4
CUDA available=True, version=12.8
CUDA devices 0: NVIDIA GeForce RTX 5090

Command 

PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True \\
CUDA_VISIBLE_DEVICES=0 \\
uv run hakari-bench evaluate from-model-card \\
  --model-card config/model_cards/sbintuitions__sarashina-embedding-v2-1b.yaml \\
  --dataset "$DATASET_GROUP" --device cuda:0 --batch-size 16 --show-progress

# A second non-overlapping dataset group was run concurrently with
# CUDA_VISIBLE_DEVICES=1 (also addressed as cuda:0 inside that process).

Submitter Notes

  • Evaluated the standard --all scope: 538 tasks across 47 datasets. The model was pinned to revision 1f3408afaa7b617e3445d891310a9c26dd0c68a5, with bf16, SDPA, and its configured 8192-token maximum sequence length.
  • Used the model-author retrieval prefixes: query task: 質問を与えるので、その質問に答えるのに役立つ関連文書を検索してください。\nquery: and document text: . No truncate/Matryoshka variant was requested because the model card does not document truncation support. The standard full-dimension int8/binary and rescore variants are included.
  • An initial batch-size-32 attempt encountered memory pressure on long documents. The canonical workers were restarted at batch size 16; 22 early files retain batch size 8 and one completed smoke result retains batch size 32. These settings do not alter the model, prompts, or sequence length.
  • The model is released under the Sarashina Model NonCommercial License Agreement; these results do not imply commercial-use permission.

Checklist

  • Result files are submitted under hakari-results/sbintuitions__sarashina-embedding-v2-1b/.
  • Result files are compressed .json.xz; no caches, DuckDB files, HTML reports, or local scratch artifacts are included.
  • The result JSON records model revision, dataset revision, runtime configuration, and package versions.
  • Overall nDCG@10 above was generated from the submitted result files.
  • Any non-default prompt, sequence length, attention implementation, candidate ranking, or reranker setting is documented above.
hotchpotch
 HAKARI-Bench org

Closing this PR because the results have already been incorporated into main through another workflow.
hotchpotch changed pull request status to closed

Sign up or log in to comment