Hugging Face's logo

Datasets:
hakari-bench
/
results

Dataset card Files
xet
 Community
37

Add results for hotchpotch/bekko-static-embedding-multilingual-v1

#37
by hotchpotch - opened
base: refs/heads/main
←
from: refs/pr/37
Discussion Files changed
+1689
-0
Add results for hotchpotch/bekko-static-embedding-multilingual-v15776b551
hotchpotch
 HAKARI-Bench org

Add HAKARI-Bench results for hotchpotch/bekko-static-embedding-multilingual-v1

Summary

Field Value
Model hotchpotch/bekko-static-embedding-multilingual-v1
Result directory hotchpotch__bekko-static-embedding-multilingual-v1
Target path hakari-results/hotchpotch__bekko-static-embedding-multilingual-v1
Result files 563 total, 563 .json.xz
Evaluation method dense
Overall nDCG@10 0.4682
Overall score units 381 grouped units from 550 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 hotchpotch/bekko-static-embedding-multilingual-v1 (1024 dims) 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 0.4682 0.5944 0.6296 0.5820 0.5149 0.4806
NanoMMTEB-v2 0.4330 0.5581 0.5590 0.4846 0.4455 0.4550
NanoRTEB 0.3750 0.6713 0.7005 0.5365 0.4711 0.3553
MNanoBEIR 0.4376 0.5509 0.6077 0.5575 0.5117 0.4646
NanoBIRCO 0.2340 0.3070 0.3526 0.2617 0.1613 0.2693
NanoMLDR 0.5572 0.6239 0.5384 0.6621 0.3920 0.7396
NanoLongEmbed 0.6196 0.7232 0.6680 0.6527 0.5014 0.8217
NanoDAPFAM 0.2053 0.3018 0.3179 0.2406 0.2380 0.2400
NanoCoIR 0.5819 0.8601 0.8777 0.6924 0.6915 0.5436
NanoIFIR 0.2383 0.3364 0.3893 0.2391 0.2152 0.2761
NanoLaw 0.5354 0.6075 0.6370 0.5597 0.4790 0.6854
NanoMedical 0.3756 0.5694 0.5803 0.5371 0.5055 0.4145
NanoRARb 0.1481 0.2689 0.2889 0.2343 0.2240 0.1359
NanoBRIGHT 0.2638 0.3885 0.4284 0.2941 0.1758 0.2790
NanoCodeRAG 0.5450 0.8712 0.9139 0.7155 0.7464 0.5823
NanoChemTEB 0.6398 0.8035 0.7980 0.7777 0.8081 0.7012
NanoR2MED 0.2173 0.3180 0.3630 0.2088 0.1099 0.2094
NanoBuiltBench 0.3989 0.5129 0.5277 0.4248 0.4291 0.3958
NanoCMTEB 0.5504 0.7982 0.8052 0.7591 0.6999 0.6003
NanoIndicQA 0.3535 0.6413 0.7056 0.7586 0.7009 0.5653
NanoMuPLeR 0.7679 0.7122 0.8388 0.8912 0.7837 0.7994
NanoMTEB-v2 0.5062 0.6372 0.6450 0.5726 0.5348 0.5028
NanoMTEB-Dutch 0.4779 0.5686 0.6213 0.5863 0.5287 0.4673
NanoMTEB-French 0.4798 0.5771 0.6377 0.5527 0.4702 0.4261
NanoMTEB-German 0.5681 0.6298 0.6536 0.6189 0.5711 0.5522
NanoJMTEB-v2 0.6536 0.7732 0.8008 0.7906 0.7165 0.7465
NanoMTEB-Korean 0.6558 0.7792 0.8246 0.8183 0.7668 0.6743
NanoFaMTEB-v2 0.5479 0.6338 0.6882 0.6652 0.6135 0.5651
NanoMTEB-Polish 0.3335 0.4738 0.5316 0.4999 0.4365 0.3424
NanoMTEB-BR 0.4951 0.6247 0.6595 0.6540 0.5224 0.5178
NanoSSRB 0.2833 0.3464 0.4340 0.2665 0.2511 0.2821
NanoRuMTEB 0.7885 0.8622 0.9121 0.9169 0.8643 0.7089
NanoMTEB-Scandinavian 0.6712 0.6981 0.7596 0.7740 0.7029 0.6091
NanoMTEB-Spanish 0.4540 0.5662 0.6292 0.5624 0.4848 0.3679
NanoMTEB-Thai 0.6477 0.7455 0.7670 0.7672 0.7107 0.5216
NanoVNMTEB 0.4126 0.5717 0.6066 0.5616 0.5197 0.4571
NanoMTEB-Misc 0.6404 0.7629 0.8011 0.7766 0.6423 0.4939
NanoMIRACL 0.6803 0.7879 0.8351 0.8475 0.7871 0.5715

Overall nDCG@10

Overall component nDCG@10 Score units Raw task results
NanoMMTEB-v2 0.4330 18 18
NanoRTEB 0.3750 14 14
MNanoBEIR 0.4376 13 182
NanoBIRCO 0.2340 5 5
NanoMLDR 0.5572 13 13
NanoLongEmbed 0.6196 6 6
NanoDAPFAM 0.2053 12 12
NanoCoIR 0.5819 10 10
NanoIFIR 0.2383 4 4
NanoLaw 0.5354 4 4
NanoMedical 0.3756 7 7
NanoRARb 0.1481 14 14
NanoBRIGHT 0.2638 20 20
NanoCodeRAG 0.5450 4 4
NanoChemTEB 0.6398 3 3
NanoR2MED 0.2173 8 8
NanoBuiltBench 0.3989 2 2
NanoCMTEB 0.5504 8 8
NanoIndicQA 0.3535 11 11
NanoMuPLeR 0.7679 14 14
NanoMTEB-v2 0.5062 10 10
NanoMTEB-Dutch 0.4779 27 27
NanoMTEB-French 0.4798 8 8
NanoMTEB-German 0.5681 5 5
NanoJMTEB-v2 0.6536 11 11
NanoMTEB-Korean 0.6558 5 5
NanoFaMTEB-v2 0.5479 17 17
NanoMTEB-Polish 0.3335 14 14
NanoMTEB-BR 0.4951 6 6
NanoSSRB 0.2833 6 6
NanoRuMTEB 0.7885 3 3
NanoMTEB-Scandinavian 0.6712 7 7
NanoMTEB-Spanish 0.4540 7 7
NanoMTEB-Thai 0.6477 9 9
NanoVNMTEB 0.4126 26 26
NanoMTEB-Misc 0.6404 12 12
NanoMIRACL 0.6803 18 18

Reproducibility

Field Value
Model source hotchpotch/bekko-static-embedding-multilingual-v1
Model revision 48e94648fd2e28dfb9ac57e8759c559fafa836d6
Dataset revision(s) 00541a0fce4048057fb7ddec30d37155a5c23d95, 01736efbaa96f020c2a4d996efdacc18071e2fcb, 017849a95097eea984680cbab35972f8d3812376, 0f3a6f43b8a26a9b8c8d5f31b09bd60dc4cd572d, 1726763179e1e114ad9ffcdc7262923471e8ecc8, ... (50 total)
Evaluated at UTC 2026-08-29T01:55:26.080958+00:00 to 2026-08-29T02:22:57.976788+00:00
Generated at UTC 2026-08-29T01:55:26.265089+00:00 to 2026-08-29T02:22:57.976810+00:00
dtype fp32
device cuda:0
batch size 4096
attention implementation not recorded
trust remote code False
max sequence length inf
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 5.14.1
sentence-transformers 5.6.0
datasets 4.8.4
CUDA available=True, version=12.8
CUDA devices 0: NVIDIA GeForce RTX 5090

Command 

CUDA_VISIBLE_DEVICES=0 \
  tmp/.venv_eval_bekko_static/bin/hakari-bench evaluate from-model-card \
  --model-card config/model_cards/hotchpotch__bekko-static-embedding-multilingual-v1.yaml \
  --all \
  --batch-size 4096 \
  --device cuda:0

The workload was executed as disjoint dataset and split shards across two
RTX 5090 GPUs. Each task used the same model card, batch size, and visible
cuda:0 device configuration shown above.

Submitter Notes

  • The checkpoint defines empty query and document prompts, so no retrieval
    prefix was added.
  • This is a SentenceTransformers StaticEmbedding model. It has no transformer
    attention implementation, and evaluation used the checkpoint's native fp32
    embedding table without trust_remote_code.
  • The native 1,024-dimensional output is represented by base. The documented
    32, 64, 128, 256, and 512 truncate dimensions are included, together with
    the standard int8, binary, and float-rescore variants.
  • Batch size 4096 was used throughout. There were no failed or omitted standard
    tasks and no sequence-length override.
  • These are complete standard --all results: 563 evaluation artifacts and
    550 canonical Overall tasks after suite duplicate exclusions.

Checklist

  • Result files are committed under hakari-results/hotchpotch__bekko-static-embedding-multilingual-v1/.
  • 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