quantumly commited on
Commit
1bde3e5
·
verified ·
1 Parent(s): 0bb3381

v0.5 LLM-features appraiser: 2026-04-27

Browse files
Files changed (1) hide show
  1. v0_5_metadata.json +444 -0
v0_5_metadata.json ADDED
@@ -0,0 +1,444 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trained_at": "2026-04-27T06:37:23.331946+00:00",
3
+ "data_run_date": "2026-04-25",
4
+ "llm_run_date": "2026-04-26",
5
+ "version": "v0.5-llm-features",
6
+ "description": "v0.4 + LLM-derived features (name_kind, cultural_origin, fame_score, crypto_relevance, brand_collision_risk, description_emb_pca32)",
7
+ "parent_version": "v0.4",
8
+ "embedders": [
9
+ "mpnet-finetuned (from v0.3)",
10
+ "BAAI/bge-base-en-v1.5"
11
+ ],
12
+ "splits": {
13
+ "train": {
14
+ "rows": 265240,
15
+ "start": "2022-01-28",
16
+ "end": "2023-09-30"
17
+ },
18
+ "val": {
19
+ "rows": 3545,
20
+ "start": "2023-10-01",
21
+ "end": "2023-12-31"
22
+ },
23
+ "test": {
24
+ "rows": 2744,
25
+ "start": "2024-01-01",
26
+ "end": "2024-05-04"
27
+ }
28
+ },
29
+ "feature_count": 212,
30
+ "n_llm_features": 56,
31
+ "feature_cols": [
32
+ "pca_000",
33
+ "pca_001",
34
+ "pca_002",
35
+ "pca_003",
36
+ "pca_004",
37
+ "pca_005",
38
+ "pca_006",
39
+ "pca_007",
40
+ "pca_008",
41
+ "pca_009",
42
+ "pca_010",
43
+ "pca_011",
44
+ "pca_012",
45
+ "pca_013",
46
+ "pca_014",
47
+ "pca_015",
48
+ "pca_016",
49
+ "pca_017",
50
+ "pca_018",
51
+ "pca_019",
52
+ "pca_020",
53
+ "pca_021",
54
+ "pca_022",
55
+ "pca_023",
56
+ "pca_024",
57
+ "pca_025",
58
+ "pca_026",
59
+ "pca_027",
60
+ "pca_028",
61
+ "pca_029",
62
+ "pca_030",
63
+ "pca_031",
64
+ "pca_032",
65
+ "pca_033",
66
+ "pca_034",
67
+ "pca_035",
68
+ "pca_036",
69
+ "pca_037",
70
+ "pca_038",
71
+ "pca_039",
72
+ "pca_040",
73
+ "pca_041",
74
+ "pca_042",
75
+ "pca_043",
76
+ "pca_044",
77
+ "pca_045",
78
+ "pca_046",
79
+ "pca_047",
80
+ "pca_048",
81
+ "pca_049",
82
+ "pca_050",
83
+ "pca_051",
84
+ "pca_052",
85
+ "pca_053",
86
+ "pca_054",
87
+ "pca_055",
88
+ "pca_056",
89
+ "pca_057",
90
+ "pca_058",
91
+ "pca_059",
92
+ "pca_060",
93
+ "pca_061",
94
+ "pca_062",
95
+ "pca_063",
96
+ "len",
97
+ "n_digits",
98
+ "n_letters",
99
+ "n_special",
100
+ "n_lower",
101
+ "n_upper",
102
+ "is_palindrome",
103
+ "is_all_digits",
104
+ "is_all_letters",
105
+ "is_ascii",
106
+ "has_unicode",
107
+ "starts_digit",
108
+ "ends_digit",
109
+ "max_char_run",
110
+ "n_unique_chars",
111
+ "in_wikipedia",
112
+ "in_geonames",
113
+ "in_us_firstname",
114
+ "in_iso3166",
115
+ "in_ticker",
116
+ "in_sec_edgar",
117
+ "in_wiktionary_en",
118
+ "wordlist_hits",
119
+ "club__prepunk_full_rankings",
120
+ "club__common_english",
121
+ "club__common_animals",
122
+ "club__gamertags",
123
+ "club__firstnames_usa",
124
+ "club__top_nouns",
125
+ "club__us_government",
126
+ "club__pokemon_gen1",
127
+ "club__catholicism",
128
+ "club__gamertags_double",
129
+ "club__wikidata_top_fantasy_char",
130
+ "club__social_handles",
131
+ "club__pokemon_gen2",
132
+ "club__top500_cities_global",
133
+ "club__top500_cities_usa",
134
+ "club__mythical_creatures",
135
+ "club__conspiracy_theories",
136
+ "club__crypto_terms",
137
+ "club__natural_wonders",
138
+ "club__pokemon_gen3",
139
+ "club__familynames_usa",
140
+ "club__country_codes",
141
+ "club__sports",
142
+ "club__top_crypto_names",
143
+ "club__logistics",
144
+ "club__home",
145
+ "club__historic_figures",
146
+ "club__holidays",
147
+ "club__top_crypto_tickers",
148
+ "club__paranormal",
149
+ "club__performing_arts",
150
+ "club__fine_art",
151
+ "club__currency_symbols",
152
+ "club__luxury",
153
+ "club__personas",
154
+ "club__finance_terms",
155
+ "club__currency_names",
156
+ "club__pokemon_gen4",
157
+ "club__gen_alpha",
158
+ "club__crayola_classic",
159
+ "club__memes",
160
+ "club__us_states",
161
+ "n_clubs",
162
+ "trademark_conflict",
163
+ "name_age_days",
164
+ "prior_transfer_count",
165
+ "fg_value",
166
+ "eth_tvl_usd",
167
+ "eth_stable_mcap",
168
+ "eth_dex_volume",
169
+ "nft_total_fee_usd",
170
+ "fame_score",
171
+ "crypto_relevance_ord",
172
+ "brand_collision_risk_ord",
173
+ "kind__concept",
174
+ "kind__random",
175
+ "kind__brand",
176
+ "kind__surname",
177
+ "kind__first_name",
178
+ "kind__abbreviation",
179
+ "kind__place",
180
+ "kind__other",
181
+ "kind__unknown",
182
+ "origin__english",
183
+ "origin__none",
184
+ "origin__mixed",
185
+ "origin__spanish",
186
+ "origin__german",
187
+ "origin__french",
188
+ "origin__japanese",
189
+ "origin__chinese",
190
+ "origin__italian",
191
+ "origin__slavic",
192
+ "origin__korean",
193
+ "origin__arabic",
194
+ "origin__other",
195
+ "origin__unknown",
196
+ "desc_pca_000",
197
+ "desc_pca_001",
198
+ "desc_pca_002",
199
+ "desc_pca_003",
200
+ "desc_pca_004",
201
+ "desc_pca_005",
202
+ "desc_pca_006",
203
+ "desc_pca_007",
204
+ "desc_pca_008",
205
+ "desc_pca_009",
206
+ "desc_pca_010",
207
+ "desc_pca_011",
208
+ "desc_pca_012",
209
+ "desc_pca_013",
210
+ "desc_pca_014",
211
+ "desc_pca_015",
212
+ "desc_pca_016",
213
+ "desc_pca_017",
214
+ "desc_pca_018",
215
+ "desc_pca_019",
216
+ "desc_pca_020",
217
+ "desc_pca_021",
218
+ "desc_pca_022",
219
+ "desc_pca_023",
220
+ "desc_pca_024",
221
+ "desc_pca_025",
222
+ "desc_pca_026",
223
+ "desc_pca_027",
224
+ "desc_pca_028",
225
+ "desc_pca_029",
226
+ "desc_pca_030",
227
+ "desc_pca_031",
228
+ "knnmp_count",
229
+ "knnmp_mean_log",
230
+ "knnmp_median_log",
231
+ "knnmp_p90_log",
232
+ "knnmp_max_sim",
233
+ "knnmp_min_sim",
234
+ "knnmp_log_max",
235
+ "knnmp_log_min",
236
+ "knnbg_count",
237
+ "knnbg_mean_log",
238
+ "knnbg_median_log",
239
+ "knnbg_p90_log",
240
+ "knnbg_max_sim",
241
+ "knnbg_min_sim",
242
+ "knnbg_log_max",
243
+ "knnbg_log_min"
244
+ ],
245
+ "pca_dim_concat": 64,
246
+ "pca_dim_description": 32,
247
+ "name_kind_values": [
248
+ "concept",
249
+ "random",
250
+ "brand",
251
+ "surname",
252
+ "first_name",
253
+ "abbreviation",
254
+ "place",
255
+ "other",
256
+ "unknown"
257
+ ],
258
+ "cultural_origin_values": [
259
+ "english",
260
+ "none",
261
+ "mixed",
262
+ "spanish",
263
+ "german",
264
+ "french",
265
+ "japanese",
266
+ "chinese",
267
+ "italian",
268
+ "slavic",
269
+ "korean",
270
+ "arabic",
271
+ "other",
272
+ "unknown"
273
+ ],
274
+ "best_iteration": 712,
275
+ "xgb_params": {
276
+ "objective": "reg:squarederror",
277
+ "eval_metric": "rmse",
278
+ "tree_method": "hist",
279
+ "device": "cuda",
280
+ "max_depth": 7,
281
+ "learning_rate": 0.04,
282
+ "subsample": 0.85,
283
+ "colsample_bytree": 0.65,
284
+ "min_child_weight": 8,
285
+ "reg_alpha": 0.5,
286
+ "reg_lambda": 2.0,
287
+ "seed": 42
288
+ },
289
+ "metrics": {
290
+ "train": {
291
+ "r2_log": 0.8462194204330444,
292
+ "rmse_log": 0.6332300305366516,
293
+ "mae_log": 0.4247952699661255,
294
+ "median_ape": 0.26942259073257446,
295
+ "bias_log": -0.0002484263095539063
296
+ },
297
+ "val": {
298
+ "r2_log": 0.6880427002906799,
299
+ "rmse_log": 1.0231839418411255,
300
+ "mae_log": 0.7111709713935852,
301
+ "median_ape": 0.4590761661529541,
302
+ "bias_log": 0.11442188918590546
303
+ },
304
+ "test": {
305
+ "r2_log": 0.41624486446380615,
306
+ "rmse_log": 1.420867681503296,
307
+ "mae_log": 1.1492146253585815,
308
+ "median_ape": 1.0128521919250488,
309
+ "bias_log": 0.6260586380958557
310
+ }
311
+ },
312
+ "top_features": [
313
+ {
314
+ "name": "knnmp_mean_log",
315
+ "gain": 2001.844970703125
316
+ },
317
+ {
318
+ "name": "knnmp_median_log",
319
+ "gain": 1155.0853271484375
320
+ },
321
+ {
322
+ "name": "knnmp_p90_log",
323
+ "gain": 832.7803344726562
324
+ },
325
+ {
326
+ "name": "is_all_digits",
327
+ "gain": 782.1705322265625
328
+ },
329
+ {
330
+ "name": "in_wikipedia",
331
+ "gain": 531.300048828125
332
+ },
333
+ {
334
+ "name": "len",
335
+ "gain": 489.9974365234375
336
+ },
337
+ {
338
+ "name": "knnmp_log_max",
339
+ "gain": 330.5997619628906
340
+ },
341
+ {
342
+ "name": "n_clubs",
343
+ "gain": 207.94166564941406
344
+ },
345
+ {
346
+ "name": "n_letters",
347
+ "gain": 196.677490234375
348
+ },
349
+ {
350
+ "name": "origin__chinese",
351
+ "gain": 170.69927978515625
352
+ },
353
+ {
354
+ "name": "ends_digit",
355
+ "gain": 165.2652587890625
356
+ },
357
+ {
358
+ "name": "knnmp_count",
359
+ "gain": 161.44723510742188
360
+ },
361
+ {
362
+ "name": "n_digits",
363
+ "gain": 159.4886016845703
364
+ },
365
+ {
366
+ "name": "name_age_days",
367
+ "gain": 156.08135986328125
368
+ },
369
+ {
370
+ "name": "n_lower",
371
+ "gain": 148.55072021484375
372
+ },
373
+ {
374
+ "name": "n_special",
375
+ "gain": 141.7484130859375
376
+ },
377
+ {
378
+ "name": "is_palindrome",
379
+ "gain": 139.62179565429688
380
+ },
381
+ {
382
+ "name": "n_unique_chars",
383
+ "gain": 124.55799102783203
384
+ },
385
+ {
386
+ "name": "eth_stable_mcap",
387
+ "gain": 124.44175720214844
388
+ },
389
+ {
390
+ "name": "trademark_conflict",
391
+ "gain": 116.59305572509766
392
+ },
393
+ {
394
+ "name": "eth_tvl_usd",
395
+ "gain": 107.47349548339844
396
+ },
397
+ {
398
+ "name": "has_unicode",
399
+ "gain": 106.94119262695312
400
+ },
401
+ {
402
+ "name": "fame_score",
403
+ "gain": 103.080078125
404
+ },
405
+ {
406
+ "name": "knnbg_log_max",
407
+ "gain": 102.78841400146484
408
+ },
409
+ {
410
+ "name": "club__prepunk_full_rankings",
411
+ "gain": 97.14898681640625
412
+ },
413
+ {
414
+ "name": "is_ascii",
415
+ "gain": 96.375
416
+ },
417
+ {
418
+ "name": "club__gamertags",
419
+ "gain": 94.76091766357422
420
+ },
421
+ {
422
+ "name": "pca_045",
423
+ "gain": 86.68685913085938
424
+ },
425
+ {
426
+ "name": "pca_002",
427
+ "gain": 85.17119598388672
428
+ },
429
+ {
430
+ "name": "pca_004",
431
+ "gain": 81.07220458984375
432
+ }
433
+ ],
434
+ "gain_split": {
435
+ "mpnet_ft_knn": 4593.555641174316,
436
+ "bge_knn": 324.30273818969727,
437
+ "llm_name_kind": 197.1177396774292,
438
+ "llm_cultural_origin": 421.92126178741455,
439
+ "llm_scores": 103.080078125,
440
+ "llm_description": 748.444676399231,
441
+ "llm_total": 1470.5637559890747
442
+ },
443
+ "wandb_run": "https://wandb.ai/quantumly-aletheia-research/ens-appraiser/runs/m1mlf4h0"
444
+ }