hotchpotch commited on
Commit
7b60dfe
·
verified ·
1 Parent(s): 2c640b1
README.md CHANGED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - multilingual
4
+ library_name: sentence-transformers
5
+ pipeline_tag: sentence-similarity
6
+ tags:
7
+ - sentence-transformers
8
+ - embeddings
9
+ - modernbert
10
+ - matryoshka
11
+ - multilingual
12
+ - retrieval
13
+ ---
14
+
15
+ # bekko-embedding-v1-a8m-pt
16
+
17
+ Pretrained checkpoint for the bekko embedding v1 a8m release candidate. This checkpoint is the pre-fine-tuning model and is mainly useful as a base model for downstream fine-tuning or ablation work.
18
+
19
+ This upload is based on the UniR v9 QAT pretraining run:
20
+
21
+ `mmB-L4-bs8192-v1-20260301-QAT-unir_v9-full-sharedid-timing/20260504_183724/final`
22
+
23
+ ## Model Details
24
+
25
+ - Architecture: ModernBERT-style encoder, 4 layers, hidden size 384
26
+ - Pooling: mean pooling
27
+ - Embedding dimension: 384
28
+ - Maximum sequence length: 8192 tokens
29
+ - Supported truncate dimensions: 384, 256, 128, 64
30
+
31
+ The model was trained with Matryoshka representation learning, so embeddings can be truncated to the supported dimensions above. Use 384 dimensions when you need the full representation, and use 256, 128, or 64 dimensions for smaller indexes or faster retrieval.
32
+
33
+ ## Intended Use
34
+
35
+ This is the pre-fine-tuning checkpoint. For most retrieval use cases, prefer the fine-tuned model:
36
+
37
+ `hotchpotch/bekko-embedding-v1-a8m`
38
+
39
+ Use this pretrain checkpoint when you need to:
40
+
41
+ - reproduce or audit the fine-tuning recipe,
42
+ - run downstream fine-tuning from the same base,
43
+ - compare pretrain vs fine-tuned behavior,
44
+ - run ablations against the UniR v9 QAT pretrain base.
45
+
46
+ ## Usage
47
+
48
+ For retrieval, prefer `encode_query()` for queries and `encode_document()` for documents. Normalize embeddings when using cosine similarity or dot-product search on normalized vectors.
49
+
50
+ ```python
51
+ from sentence_transformers import SentenceTransformer
52
+
53
+ model = SentenceTransformer("hotchpotch/bekko-embedding-v1-a8m-pt")
54
+
55
+ queries = ["What is multilingual retrieval?"]
56
+ documents = ["Multilingual retrieval searches documents across languages."]
57
+
58
+ query_embeddings = model.encode_query(queries, normalize_embeddings=True)
59
+ document_embeddings = model.encode_document(documents, normalize_embeddings=True)
60
+
61
+ scores = query_embeddings @ document_embeddings.T
62
+ ```
63
+
64
+ ## Notes
65
+
66
+ This repository is intended for private evaluation and release-candidate validation. The model configuration in this upload is set for 8192-token inference.
cli_args.json ADDED
The diff for this file is too large to render. See raw diff
 
config_sentence_transformers.json CHANGED
@@ -11,4 +11,4 @@
11
  },
12
  "default_prompt_name": null,
13
  "similarity_fn_name": "cosine"
14
- }
 
11
  },
12
  "default_prompt_name": null,
13
  "similarity_fn_name": "cosine"
14
+ }
eval_result.json ADDED
@@ -0,0 +1,274 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "nano_beir": {
3
+ "NanoArguAna": {
4
+ "MNanoArguAna-ar_cosine_ndcg@10": 0.4730338404204377,
5
+ "MNanoArguAna-de_cosine_ndcg@10": 0.4523858022833368,
6
+ "MNanoArguAna-en_cosine_ndcg@10": 0.5593102661290839,
7
+ "MNanoArguAna-es_cosine_ndcg@10": 0.5237835854585444,
8
+ "MNanoArguAna-fr_cosine_ndcg@10": 0.5168055525240536,
9
+ "MNanoArguAna-it_cosine_ndcg@10": 0.4727147681288612,
10
+ "MNanoArguAna-mean_cosine_ndcg@10": 0.49946905572549927,
11
+ "MNanoArguAna-no_cosine_ndcg@10": 0.5138751897026443,
12
+ "MNanoArguAna-pt_cosine_ndcg@10": 0.5034692240128686,
13
+ "MNanoArguAna-sv_cosine_ndcg@10": 0.4798432728696626
14
+ },
15
+ "NanoClimateFEVER": {
16
+ "MNanoClimateFEVER-ar_cosine_ndcg@10": 0.21696059280899108,
17
+ "MNanoClimateFEVER-de_cosine_ndcg@10": 0.28253387055871765,
18
+ "MNanoClimateFEVER-en_cosine_ndcg@10": 0.2916839711692681,
19
+ "MNanoClimateFEVER-es_cosine_ndcg@10": 0.3448162987069231,
20
+ "MNanoClimateFEVER-fr_cosine_ndcg@10": 0.22376330656871984,
21
+ "MNanoClimateFEVER-it_cosine_ndcg@10": 0.2782216799184051,
22
+ "MNanoClimateFEVER-mean_cosine_ndcg@10": 0.2762952444548678,
23
+ "MNanoClimateFEVER-no_cosine_ndcg@10": 0.27737273672960566,
24
+ "MNanoClimateFEVER-pt_cosine_ndcg@10": 0.29771166786184095,
25
+ "MNanoClimateFEVER-sv_cosine_ndcg@10": 0.27359307577133846
26
+ },
27
+ "NanoClimateFEVER-ja": {
28
+ "MNanoClimateFEVER-ja-mean_cosine_ndcg@10": 0.35233710769938187,
29
+ "MNanoClimateFEVER-ja_cosine_ndcg@10": 0.35233710769938187
30
+ },
31
+ "NanoDBPedia": {
32
+ "MNanoDBPedia-ar_cosine_ndcg@10": 0.49176953708703963,
33
+ "MNanoDBPedia-de_cosine_ndcg@10": 0.5585728125958513,
34
+ "MNanoDBPedia-en_cosine_ndcg@10": 0.6013656077919083,
35
+ "MNanoDBPedia-es_cosine_ndcg@10": 0.5635691591790805,
36
+ "MNanoDBPedia-fr_cosine_ndcg@10": 0.5598146452043108,
37
+ "MNanoDBPedia-it_cosine_ndcg@10": 0.5786410295503902,
38
+ "MNanoDBPedia-mean_cosine_ndcg@10": 0.5578189519284785,
39
+ "MNanoDBPedia-no_cosine_ndcg@10": 0.5571766709488738,
40
+ "MNanoDBPedia-pt_cosine_ndcg@10": 0.5449366078473445,
41
+ "MNanoDBPedia-sv_cosine_ndcg@10": 0.5645244971515071
42
+ },
43
+ "NanoDBPedia-ja": {
44
+ "MNanoDBPedia-ja-mean_cosine_ndcg@10": 0.5792733395527497,
45
+ "MNanoDBPedia-ja_cosine_ndcg@10": 0.5792733395527497
46
+ },
47
+ "NanoFEVER": {
48
+ "MNanoFEVER-ar_cosine_ndcg@10": 0.7811365282748113,
49
+ "MNanoFEVER-de_cosine_ndcg@10": 0.7907331358379059,
50
+ "MNanoFEVER-en_cosine_ndcg@10": 0.8609922099824149,
51
+ "MNanoFEVER-es_cosine_ndcg@10": 0.8338478791796845,
52
+ "MNanoFEVER-fr_cosine_ndcg@10": 0.6787773582711195,
53
+ "MNanoFEVER-it_cosine_ndcg@10": 0.8033124773204208,
54
+ "MNanoFEVER-mean_cosine_ndcg@10": 0.7977775878693328,
55
+ "MNanoFEVER-no_cosine_ndcg@10": 0.7903408907064573,
56
+ "MNanoFEVER-pt_cosine_ndcg@10": 0.8221987118236579,
57
+ "MNanoFEVER-sv_cosine_ndcg@10": 0.8186590994275242
58
+ },
59
+ "NanoFEVER-ja": {
60
+ "MNanoFEVER-ja-mean_cosine_ndcg@10": 0.724125971143851,
61
+ "MNanoFEVER-ja_cosine_ndcg@10": 0.724125971143851
62
+ },
63
+ "NanoFiQA2018": {
64
+ "MNanoFiQA2018-ar_cosine_ndcg@10": 0.32000050872745656,
65
+ "MNanoFiQA2018-de_cosine_ndcg@10": 0.33859658725536546,
66
+ "MNanoFiQA2018-en_cosine_ndcg@10": 0.45818623330557856,
67
+ "MNanoFiQA2018-es_cosine_ndcg@10": 0.33427309933119126,
68
+ "MNanoFiQA2018-fr_cosine_ndcg@10": 0.3092582897093253,
69
+ "MNanoFiQA2018-it_cosine_ndcg@10": 0.33185375268271744,
70
+ "MNanoFiQA2018-mean_cosine_ndcg@10": 0.3243882573066615,
71
+ "MNanoFiQA2018-no_cosine_ndcg@10": 0.29262339645711927,
72
+ "MNanoFiQA2018-pt_cosine_ndcg@10": 0.2682571114141015,
73
+ "MNanoFiQA2018-sv_cosine_ndcg@10": 0.2664453368770985
74
+ },
75
+ "NanoFiQA2018-ja": {
76
+ "MNanoFiQA2018-ja-mean_cosine_ndcg@10": 0.3222650200367273,
77
+ "MNanoFiQA2018-ja_cosine_ndcg@10": 0.3222650200367273
78
+ },
79
+ "NanoHotpotQA": {
80
+ "MNanoHotpotQA-ar_cosine_ndcg@10": 0.6956427737846929,
81
+ "MNanoHotpotQA-de_cosine_ndcg@10": 0.7701008639893859,
82
+ "MNanoHotpotQA-en_cosine_ndcg@10": 0.7690504439859169,
83
+ "MNanoHotpotQA-es_cosine_ndcg@10": 0.722497873994412,
84
+ "MNanoHotpotQA-fr_cosine_ndcg@10": 0.7463668087829665,
85
+ "MNanoHotpotQA-it_cosine_ndcg@10": 0.7464549361636561,
86
+ "MNanoHotpotQA-mean_cosine_ndcg@10": 0.7416132572210601,
87
+ "MNanoHotpotQA-no_cosine_ndcg@10": 0.7401936856204756,
88
+ "MNanoHotpotQA-pt_cosine_ndcg@10": 0.734274156218506,
89
+ "MNanoHotpotQA-sv_cosine_ndcg@10": 0.7499377724495281
90
+ },
91
+ "NanoHotpotQA-ja": {
92
+ "MNanoHotpotQA-ja-mean_cosine_ndcg@10": 0.6701411332471646,
93
+ "MNanoHotpotQA-ja_cosine_ndcg@10": 0.6701411332471646
94
+ },
95
+ "NanoMSMARCO": {
96
+ "MNanoMSMARCO-ar_cosine_ndcg@10": 0.43785211849573136,
97
+ "MNanoMSMARCO-de_cosine_ndcg@10": 0.47974421951438656,
98
+ "MNanoMSMARCO-en_cosine_ndcg@10": 0.5476519075374822,
99
+ "MNanoMSMARCO-es_cosine_ndcg@10": 0.49934336391420997,
100
+ "MNanoMSMARCO-fr_cosine_ndcg@10": 0.4821488420046777,
101
+ "MNanoMSMARCO-it_cosine_ndcg@10": 0.5089587403780814,
102
+ "MNanoMSMARCO-mean_cosine_ndcg@10": 0.4732463048264413,
103
+ "MNanoMSMARCO-no_cosine_ndcg@10": 0.4077398253644848,
104
+ "MNanoMSMARCO-pt_cosine_ndcg@10": 0.49901980528206363,
105
+ "MNanoMSMARCO-sv_cosine_ndcg@10": 0.39675792094685436
106
+ },
107
+ "NanoMSMARCO-ja": {
108
+ "MNanoMSMARCO-ja-mean_cosine_ndcg@10": 0.4945060711606406,
109
+ "MNanoMSMARCO-ja_cosine_ndcg@10": 0.4945060711606406
110
+ },
111
+ "NanoNFCorpus": {
112
+ "MNanoNFCorpus-ar_cosine_ndcg@10": 0.22716153322345686,
113
+ "MNanoNFCorpus-de_cosine_ndcg@10": 0.24481731941369972,
114
+ "MNanoNFCorpus-en_cosine_ndcg@10": 0.33595088460224604,
115
+ "MNanoNFCorpus-es_cosine_ndcg@10": 0.30753131265305245,
116
+ "MNanoNFCorpus-fr_cosine_ndcg@10": 0.3188369766309311,
117
+ "MNanoNFCorpus-it_cosine_ndcg@10": 0.28557450822422525,
118
+ "MNanoNFCorpus-mean_cosine_ndcg@10": 0.28659175055370834,
119
+ "MNanoNFCorpus-no_cosine_ndcg@10": 0.2725952012643174,
120
+ "MNanoNFCorpus-pt_cosine_ndcg@10": 0.2979627431402395,
121
+ "MNanoNFCorpus-sv_cosine_ndcg@10": 0.2888952758312067
122
+ },
123
+ "NanoNFCorpus-ja": {
124
+ "MNanoNFCorpus-ja-mean_cosine_ndcg@10": 0.2799070812314455,
125
+ "MNanoNFCorpus-ja_cosine_ndcg@10": 0.2799070812314455
126
+ },
127
+ "NanoNQ": {
128
+ "MNanoNQ-ar_cosine_ndcg@10": 0.45515849206895986,
129
+ "MNanoNQ-de_cosine_ndcg@10": 0.49779262972522065,
130
+ "MNanoNQ-en_cosine_ndcg@10": 0.6090497523222711,
131
+ "MNanoNQ-es_cosine_ndcg@10": 0.43973148585510474,
132
+ "MNanoNQ-fr_cosine_ndcg@10": 0.5161833937286435,
133
+ "MNanoNQ-it_cosine_ndcg@10": 0.4683263144521548,
134
+ "MNanoNQ-mean_cosine_ndcg@10": 0.47901010889360535,
135
+ "MNanoNQ-no_cosine_ndcg@10": 0.44336355209988704,
136
+ "MNanoNQ-pt_cosine_ndcg@10": 0.4340017129052539,
137
+ "MNanoNQ-sv_cosine_ndcg@10": 0.44748364688495246
138
+ },
139
+ "NanoNQ-ja": {
140
+ "MNanoNQ-ja-mean_cosine_ndcg@10": 0.5403929314434924,
141
+ "MNanoNQ-ja_cosine_ndcg@10": 0.5403929314434924
142
+ },
143
+ "NanoQuoraRetrieval": {
144
+ "MNanoQuoraRetrieval-ar_cosine_ndcg@10": 0.8180530918417213,
145
+ "MNanoQuoraRetrieval-de_cosine_ndcg@10": 0.8356060159578926,
146
+ "MNanoQuoraRetrieval-en_cosine_ndcg@10": 0.9346258899873209,
147
+ "MNanoQuoraRetrieval-es_cosine_ndcg@10": 0.8821377839507355,
148
+ "MNanoQuoraRetrieval-fr_cosine_ndcg@10": 0.7925711224415832,
149
+ "MNanoQuoraRetrieval-it_cosine_ndcg@10": 0.8340771342434293,
150
+ "MNanoQuoraRetrieval-mean_cosine_ndcg@10": 0.8307165948334699,
151
+ "MNanoQuoraRetrieval-no_cosine_ndcg@10": 0.7588980586522337,
152
+ "MNanoQuoraRetrieval-pt_cosine_ndcg@10": 0.8207898164276353,
153
+ "MNanoQuoraRetrieval-sv_cosine_ndcg@10": 0.7996904399986784
154
+ },
155
+ "NanoQuoraRetrieval-ja": {
156
+ "MNanoQuoraRetrieval-ja-mean_cosine_ndcg@10": 0.8935715249766721,
157
+ "MNanoQuoraRetrieval-ja_cosine_ndcg@10": 0.8935715249766721
158
+ },
159
+ "NanoSCIDOCS": {
160
+ "MNanoSCIDOCS-ar_cosine_ndcg@10": 0.2935560321306876,
161
+ "MNanoSCIDOCS-de_cosine_ndcg@10": 0.30095111574851746,
162
+ "MNanoSCIDOCS-en_cosine_ndcg@10": 0.3821570242399539,
163
+ "MNanoSCIDOCS-es_cosine_ndcg@10": 0.3421588969001715,
164
+ "MNanoSCIDOCS-fr_cosine_ndcg@10": 0.3082776916852559,
165
+ "MNanoSCIDOCS-it_cosine_ndcg@10": 0.3109347050091602,
166
+ "MNanoSCIDOCS-mean_cosine_ndcg@10": 0.31933584744766974,
167
+ "MNanoSCIDOCS-no_cosine_ndcg@10": 0.2924377548797874,
168
+ "MNanoSCIDOCS-pt_cosine_ndcg@10": 0.3284645551969043,
169
+ "MNanoSCIDOCS-sv_cosine_ndcg@10": 0.31508485123858976
170
+ },
171
+ "NanoSCIDOCS-ja": {
172
+ "MNanoSCIDOCS-ja-mean_cosine_ndcg@10": 0.34071630056860536,
173
+ "MNanoSCIDOCS-ja_cosine_ndcg@10": 0.34071630056860536
174
+ },
175
+ "NanoSciFact": {
176
+ "MNanoSciFact-ar_cosine_ndcg@10": 0.5288426430511044,
177
+ "MNanoSciFact-de_cosine_ndcg@10": 0.6622217521928413,
178
+ "MNanoSciFact-en_cosine_ndcg@10": 0.7628440491657529,
179
+ "MNanoSciFact-es_cosine_ndcg@10": 0.6527242947213842,
180
+ "MNanoSciFact-fr_cosine_ndcg@10": 0.676593207849287,
181
+ "MNanoSciFact-it_cosine_ndcg@10": 0.6419598857336184,
182
+ "MNanoSciFact-mean_cosine_ndcg@10": 0.6344163638808986,
183
+ "MNanoSciFact-no_cosine_ndcg@10": 0.6071069309322736,
184
+ "MNanoSciFact-pt_cosine_ndcg@10": 0.5776104964850115,
185
+ "MNanoSciFact-sv_cosine_ndcg@10": 0.5998440147968133
186
+ },
187
+ "NanoTouche2020": {
188
+ "MNanoTouche2020-ar_cosine_ndcg@10": 0.4081229095460993,
189
+ "MNanoTouche2020-de_cosine_ndcg@10": 0.4388518349118278,
190
+ "MNanoTouche2020-en_cosine_ndcg@10": 0.5124116336684273,
191
+ "MNanoTouche2020-es_cosine_ndcg@10": 0.44235774385322657,
192
+ "MNanoTouche2020-fr_cosine_ndcg@10": 0.3992503328132656,
193
+ "MNanoTouche2020-it_cosine_ndcg@10": 0.4076703520501428,
194
+ "MNanoTouche2020-mean_cosine_ndcg@10": 0.4335759729678913,
195
+ "MNanoTouche2020-no_cosine_ndcg@10": 0.4241551560510449,
196
+ "MNanoTouche2020-pt_cosine_ndcg@10": 0.43129192978542286,
197
+ "MNanoTouche2020-sv_cosine_ndcg@10": 0.43807186403156484
198
+ }
199
+ },
200
+ "nano_fineweb2ir": {
201
+ "NanoFineWeb2IR-arb_Arab_cosine_ndcg@10": 0.9438507213043262,
202
+ "NanoFineWeb2IR-ces_Latn_cosine_ndcg@10": 0.9452371901428583,
203
+ "NanoFineWeb2IR-cmn_Hani_cosine_ndcg@10": 0.9238507213043262,
204
+ "NanoFineWeb2IR-dan_Latn_cosine_ndcg@10": 0.9578557852142875,
205
+ "NanoFineWeb2IR-deu_Latn_cosine_ndcg@10": 0.9926185950714291,
206
+ "NanoFineWeb2IR-ell_Grek_cosine_ndcg@10": 0.9406803948112681,
207
+ "NanoFineWeb2IR-eng_Latn_cosine_ndcg@10": 0.9686135311614679,
208
+ "NanoFineWeb2IR-fas_Arab_cosine_ndcg@10": 0.9287119211911774,
209
+ "NanoFineWeb2IR-fra_Latn_cosine_ndcg@10": 0.8653058616513755,
210
+ "NanoFineWeb2IR-hun_Latn_cosine_ndcg@10": 0.9332476622362154,
211
+ "NanoFineWeb2IR-ind_Latn_cosine_ndcg@10": 0.9200983900297096,
212
+ "NanoFineWeb2IR-ita_Latn_cosine_ndcg@10": 0.8565362699694403,
213
+ "NanoFineWeb2IR-jpn_Jpan_cosine_ndcg@10": 0.9926185950714291,
214
+ "NanoFineWeb2IR-nld_Latn_cosine_ndcg@10": 0.9426185950714292,
215
+ "NanoFineWeb2IR-pol_Latn_cosine_ndcg@10": 0.9278557852142875,
216
+ "NanoFineWeb2IR-por_Latn_cosine_ndcg@10": 0.9478557852142875,
217
+ "NanoFineWeb2IR-rus_Cyrl_cosine_ndcg@10": 0.96,
218
+ "NanoFineWeb2IR-spa_Latn_cosine_ndcg@10": 0.9483998915977869,
219
+ "NanoFineWeb2IR-swe_Latn_cosine_ndcg@10": 0.9852371901428583,
220
+ "NanoFineWeb2IR-tur_Latn_cosine_ndcg@10": 0.9429742462875491,
221
+ "NanoFineWeb2IR-vie_Latn_cosine_ndcg@10": 0.9167836778214311,
222
+ "NanoFineWeb2IR_mean_cosine_ndcg@10": 0.9400452766909018
223
+ },
224
+ "nano_miracl": {
225
+ "NanoMIRACL-ar_cosine_ndcg@10": 0.6955587747416526,
226
+ "NanoMIRACL-bn_cosine_ndcg@10": 0.6443383048804079,
227
+ "NanoMIRACL-de_cosine_ndcg@10": 0.6099802384119607,
228
+ "NanoMIRACL-en_cosine_ndcg@10": 0.6734205400931867,
229
+ "NanoMIRACL-es_cosine_ndcg@10": 0.6876805333372731,
230
+ "NanoMIRACL-fa_cosine_ndcg@10": 0.6484924725291072,
231
+ "NanoMIRACL-fi_cosine_ndcg@10": 0.7461712879088102,
232
+ "NanoMIRACL-fr_cosine_ndcg@10": 0.6217460185241914,
233
+ "NanoMIRACL-hi_cosine_ndcg@10": 0.6157970155676352,
234
+ "NanoMIRACL-id_cosine_ndcg@10": 0.656938278826782,
235
+ "NanoMIRACL-ja_cosine_ndcg@10": 0.6269369051853845,
236
+ "NanoMIRACL-ko_cosine_ndcg@10": 0.5944354127162638,
237
+ "NanoMIRACL-ru_cosine_ndcg@10": 0.7159646473373542,
238
+ "NanoMIRACL-sw_cosine_ndcg@10": 0.5656108922525411,
239
+ "NanoMIRACL-te_cosine_ndcg@10": 0.9386391949847088,
240
+ "NanoMIRACL-th_cosine_ndcg@10": 0.7792507958853974,
241
+ "NanoMIRACL-yo_cosine_ndcg@10": 0.6907384653898725,
242
+ "NanoMIRACL-zh_cosine_ndcg@10": 0.6344802314104463,
243
+ "NanoMIRACL_mean_cosine_ndcg@10": 0.6747877783323876
244
+ },
245
+ "summary": {
246
+ "NanoArguAna_mean_ndcg@10": 0.49946905572549927,
247
+ "NanoClimateFEVER-ja_mean_ndcg@10": 0.35233710769938187,
248
+ "NanoClimateFEVER_mean_ndcg@10": 0.2762952444548678,
249
+ "NanoDBPedia-ja_mean_ndcg@10": 0.5792733395527497,
250
+ "NanoDBPedia_mean_ndcg@10": 0.5578189519284785,
251
+ "NanoFEVER-ja_mean_ndcg@10": 0.724125971143851,
252
+ "NanoFEVER_mean_ndcg@10": 0.7977775878693328,
253
+ "NanoFiQA2018-ja_mean_ndcg@10": 0.3222650200367273,
254
+ "NanoFiQA2018_mean_ndcg@10": 0.3243882573066615,
255
+ "NanoHotpotQA-ja_mean_ndcg@10": 0.6701411332471646,
256
+ "NanoHotpotQA_mean_ndcg@10": 0.7416132572210601,
257
+ "NanoMSMARCO-ja_mean_ndcg@10": 0.4945060711606406,
258
+ "NanoMSMARCO_mean_ndcg@10": 0.4732463048264413,
259
+ "NanoNFCorpus-ja_mean_ndcg@10": 0.2799070812314455,
260
+ "NanoNFCorpus_mean_ndcg@10": 0.28659175055370834,
261
+ "NanoNQ-ja_mean_ndcg@10": 0.5403929314434924,
262
+ "NanoNQ_mean_ndcg@10": 0.47901010889360535,
263
+ "NanoQuoraRetrieval-ja_mean_ndcg@10": 0.8935715249766721,
264
+ "NanoQuoraRetrieval_mean_ndcg@10": 0.8307165948334699,
265
+ "NanoSCIDOCS-ja_mean_ndcg@10": 0.34071630056860536,
266
+ "NanoSCIDOCS_mean_ndcg@10": 0.31933584744766974,
267
+ "NanoSciFact_mean_ndcg@10": 0.6344163638808986,
268
+ "NanoTouche2020_mean_ndcg@10": 0.4335759729678913,
269
+ "all_mean_ndcg@10": 0.5386529933597441,
270
+ "nano_fineweb2ir_mean_ndcg@10": 0.9400452766909018,
271
+ "nano_miracl_mean_ndcg@10": 0.6747877783323876
272
+ },
273
+ "timestamp_utc": "2026-07-06T18:00:23.553943+00:00"
274
+ }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b4fe25becb21b90d7cca210d8446bbcc450dff2fee461c32baf41ef442dd2ef4
3
  size 211952896
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b474c13e826f1652bffef917e1e3a1dccb172607087ccd0d5a0a61dc91c83cc7
3
  size 211952896
tokenizer_config.json CHANGED
@@ -2004,12 +2004,12 @@
2004
  "eos_token": "<eos>",
2005
  "extra_special_tokens": {},
2006
  "mask_token": "<mask>",
2007
- "max_length": 512,
2008
  "model_input_names": [
2009
  "input_ids",
2010
  "attention_mask"
2011
  ],
2012
- "model_max_length": 512,
2013
  "pad_to_multiple_of": null,
2014
  "pad_token": "<pad>",
2015
  "pad_token_type_id": 0,
 
2004
  "eos_token": "<eos>",
2005
  "extra_special_tokens": {},
2006
  "mask_token": "<mask>",
2007
+ "max_length": 8192,
2008
  "model_input_names": [
2009
  "input_ids",
2010
  "attention_mask"
2011
  ],
2012
+ "model_max_length": 8192,
2013
  "pad_to_multiple_of": null,
2014
  "pad_token": "<pad>",
2015
  "pad_token_type_id": 0,