Hugging Face's logo

Datasets:
hakari-bench
/
results

Dataset card Files
xet
 Community
26

Add ST 4.0.2 results for sbintuitions/sarashina-embedding-v2-1b

#23
by hotchpotch - opened
base: refs/heads/main
←
from: refs/pr/23
Discussion Files changed
+1614
-0
Add ST 4.0.2 results for sbintuitions/sarashina-embedding-v2-1b53749bcc
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 0.4739
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 0.4739 0.5979 0.6323 0.5859 0.5190 0.4832
NanoMMTEB-v2 0.4582 0.5581 0.5590 0.4846 0.4455 0.4550
NanoRTEB 0.5961 0.6713 0.7005 0.5365 0.4711 0.3553
MNanoBEIR 0.4238 0.5509 0.6077 0.5575 0.5117 0.4646
NanoBIRCO 0.3163 0.3070 0.3526 0.2617 0.1613 0.2693
NanoMLDR 0.4326 0.6239 0.5384 0.6621 0.3920 0.7396
NanoLongEmbed 0.6163 0.7232 0.6680 0.6527 0.5014 0.8217
NanoDAPFAM 0.2835 0.3018 0.3179 0.2406 0.2380 0.2400
NanoCoIR 0.7529 0.8601 0.8777 0.6924 0.6915 0.5436
NanoIFIR 0.2996 0.3364 0.3893 0.2391 0.2152 0.2761
NanoLaw 0.4858 0.6075 0.6370 0.5597 0.4790 0.6854
NanoMedical 0.4203 0.5694 0.5803 0.5371 0.5055 0.4145
NanoRARb 0.2438 0.2689 0.2889 0.2343 0.2240 0.1359
NanoBRIGHT 0.4279 0.3885 0.4284 0.2941 0.1758 0.2790
NanoCodeRAG 0.8695 0.8712 0.9139 0.7155 0.7464 0.5823
NanoChemTEB 0.8003 0.8035 0.7980 0.7777 0.8081 0.7012
NanoR2MED 0.3254 0.3180 0.3630 0.2088 0.1099 0.2094
NanoBuiltBench 0.5407 0.5129 0.5277 0.4248 0.4291 0.3958
NanoCMTEB 0.6316 0.7982 0.8052 0.7591 0.6999 0.6003
NanoIndicQA 0.1671 0.6413 0.7056 0.7586 0.7009 0.5653
NanoMuPLeR 0.4887 0.7122 0.8388 0.8912 0.7837 0.7994
NanoMTEB-v2 0.5982 0.6372 0.6450 0.5726 0.5348 0.5028
NanoMTEB-Dutch 0.4924 0.5686 0.6213 0.5863 0.5287 0.4673
NanoMTEB-French 0.4955 0.5771 0.6377 0.5527 0.4702 0.4261
NanoMTEB-German 0.5423 0.6298 0.6536 0.6189 0.5711 0.5522
NanoJMTEB-v2 0.8303 0.7732 0.8008 0.7906 0.7165 0.7465
NanoMTEB-Korean 0.6217 0.7792 0.8246 0.8183 0.7668 0.6743
NanoFaMTEB-v2 0.3048 0.6338 0.6882 0.6652 0.6135 0.5651
NanoMTEB-Polish 0.3250 0.4738 0.5316 0.4999 0.4365 0.3424
NanoRuMTEB 0.7550 0.8622 0.9121 0.9169 0.8643 0.7089
NanoMTEB-Scandinavian 0.6069 0.6981 0.7596 0.7740 0.7029 0.6091
NanoMTEB-Spanish 0.4838 0.5662 0.6292 0.5624 0.4848 0.3679
NanoMTEB-Thai 0.3490 0.7455 0.7670 0.7672 0.7107 0.5216
NanoVNMTEB 0.4161 0.5717 0.6066 0.5616 0.5197 0.4571
NanoMTEB-Misc 0.6572 0.7629 0.8011 0.7766 0.6423 0.4939
NanoMIRACL 0.5512 0.7879 0.8351 0.8475 0.7871 0.5715

Overall nDCG@10

Overall component nDCG@10 Score units Raw task results
NanoMMTEB-v2 0.4582 18 18
NanoRTEB 0.5961 14 14
MNanoBEIR 0.4238 13 169
NanoBIRCO 0.3163 5 5
NanoMLDR 0.4326 13 13
NanoLongEmbed 0.6163 6 6
NanoDAPFAM 0.2835 12 12
NanoCoIR 0.7529 10 10
NanoIFIR 0.2996 4 4
NanoLaw 0.4858 4 4
NanoMedical 0.4203 7 7
NanoRARb 0.2438 14 14
NanoBRIGHT 0.4279 20 20
NanoCodeRAG 0.8695 4 4
NanoChemTEB 0.8003 3 3
NanoR2MED 0.3254 8 8
NanoBuiltBench 0.5407 2 2
NanoCMTEB 0.6316 8 8
NanoIndicQA 0.1671 11 11
NanoMuPLeR 0.4887 14 14
NanoMTEB-v2 0.5982 10 10
NanoMTEB-Dutch 0.4924 27 27
NanoMTEB-French 0.4955 8 8
NanoMTEB-German 0.5423 5 5
NanoJMTEB-v2 0.8303 11 11
NanoMTEB-Korean 0.6217 5 5
NanoFaMTEB-v2 0.3048 17 17
NanoMTEB-Polish 0.3250 14 14
NanoRuMTEB 0.7550 3 3
NanoMTEB-Scandinavian 0.6069 7 7
NanoMTEB-Spanish 0.4838 7 7
NanoMTEB-Thai 0.3490 9 9
NanoVNMTEB 0.4161 26 26
NanoMTEB-Misc 0.6572 12 12
NanoMIRACL 0.5512 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 2026-07-15T21:27:47.810203+00:00 to 2026-07-16T11:22:24.803640+00:00
Generated at UTC 2026-07-15T21:27:48.004270+00:00 to 2026-07-16T11:22:24.803658+00:00
dtype bf16
device cuda:0
batch size 16
attention implementation flash_attention_2
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 4.57.6
sentence-transformers 4.0.2
datasets 4.8.4
CUDA available=True, version=12.8
CUDA devices 0: NVIDIA GeForce RTX 5090

Command 

CUDA_VISIBLE_DEVICES=1 PYTHONPATH="$PWD" PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True tmp/venvs/sarashina-st402-tf4-fa2/bin/python -m hakari_bench.cli evaluate dense --model sbintuitions/sarashina-embedding-v2-1b --model-revision 1f3408afaa7b617e3445d891310a9c26dd0c68a5 --all --dtype bf16 --flash-attn2 --device cuda:0 --batch-size 16 --query-prompt "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: " --document-prompt "text: " --results-dir output/sarashina-st402-tf4-fa2-full --show-progress

Submitter Notes

  • Used the official SentenceTransformers 4.0.2-compatible runtime: Transformers 4.57.6, FlashAttention 2.8.3, bf16, batch size 16, and the model-card retrieval query/document prompts.
  • No retries, task reruns, or sequence-length changes; max sequence length remained 8192.
  • Standard --all result coverage: 538 tasks.

Checklist

  • Result files are committed 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 changed pull request status to merged

Sign up or log in to comment