argus-uq 0.1.0__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (142) hide show
  1. argus_uq/__init__.py +24 -0
  2. argus_uq/_data/derived/ause_winners.csv +17 -0
  3. argus_uq/_data/derived/ause_winners_closed.csv +13 -0
  4. argus_uq/_data/derived/bootstrap_ci.csv +2593 -0
  5. argus_uq/_data/derived/bootstrap_ci_closed.csv +577 -0
  6. argus_uq/_data/derived/cell_winners.csv +17 -0
  7. argus_uq/_data/derived/cell_winners_closed.csv +13 -0
  8. argus_uq/_data/derived/conformal_disks.csv +7201 -0
  9. argus_uq/_data/derived/conformal_disks_closed.csv +5401 -0
  10. argus_uq/_data/derived/conformal_disks_summary.csv +145 -0
  11. argus_uq/_data/derived/conformal_disks_summary_closed.csv +109 -0
  12. argus_uq/_data/derived/cross_cell_rho.csv +17 -0
  13. argus_uq/_data/derived/cross_cell_rho_closed.csv +13 -0
  14. argus_uq/_data/derived/cross_cell_rho_pairs.csv +121 -0
  15. argus_uq/_data/derived/cross_cell_rho_pairs_closed.csv +67 -0
  16. argus_uq/_data/derived/cross_cell_rho_stats.csv +2 -0
  17. argus_uq/_data/derived/cross_cell_rho_stats_closed.csv +2 -0
  18. argus_uq/_data/derived/cross_open_closed_rho.csv +13 -0
  19. argus_uq/_data/derived/family_best_per_cell.csv +113 -0
  20. argus_uq/_data/derived/family_best_per_cell_closed.csv +49 -0
  21. argus_uq/_data/derived/family_means.csv +113 -0
  22. argus_uq/_data/derived/family_means_closed.csv +49 -0
  23. argus_uq/_data/derived/method_leaderboard.csv +28 -0
  24. argus_uq/_data/derived/method_leaderboard_closed.csv +9 -0
  25. argus_uq/_data/derived/nan_audit.csv +31 -0
  26. argus_uq/_data/derived/nan_audit_closed.csv +1 -0
  27. argus_uq/_data/derived/significance_pairs.csv +49 -0
  28. argus_uq/_data/derived/significance_pairs_closed.csv +37 -0
  29. argus_uq/_data/derived/tuning_deltas.csv +28 -0
  30. argus_uq/_data/derived/tuning_deltas_closed.csv +1 -0
  31. argus_uq/_data/manifest.json +66 -0
  32. argus_uq/_data/panel_drops.json +32 -0
  33. argus_uq/_data/per_item_full/GemxOSG.parquet +0 -0
  34. argus_uq/_data/per_item_full/GemxSP.parquet +0 -0
  35. argus_uq/_data/per_item_full/GemxUIV.parquet +0 -0
  36. argus_uq/_data/per_item_full/GemxV2.parquet +0 -0
  37. argus_uq/_data/per_item_full/OAIxOSG.parquet +0 -0
  38. argus_uq/_data/per_item_full/OAIxSP.parquet +0 -0
  39. argus_uq/_data/per_item_full/OAIxUIV.parquet +0 -0
  40. argus_uq/_data/per_item_full/OAIxV2.parquet +0 -0
  41. argus_uq/_data/per_item_full/PTxOSG.parquet +0 -0
  42. argus_uq/_data/per_item_full/PTxSP.parquet +0 -0
  43. argus_uq/_data/per_item_full/PTxV2.parquet +0 -0
  44. argus_uq/_data/per_item_full/PxUIV.parquet +0 -0
  45. argus_uq/_data/per_item_full/Q72xOSG.parquet +0 -0
  46. argus_uq/_data/per_item_full/Q72xSP.parquet +0 -0
  47. argus_uq/_data/per_item_full/Q72xUIV.parquet +0 -0
  48. argus_uq/_data/per_item_full/Q72xV2.parquet +0 -0
  49. argus_uq/_data/per_item_full/Q7xOSG.parquet +0 -0
  50. argus_uq/_data/per_item_full/Q7xSP.parquet +0 -0
  51. argus_uq/_data/per_item_full/Q7xUIV.parquet +0 -0
  52. argus_uq/_data/per_item_full/Q7xV2.parquet +0 -0
  53. argus_uq/_data/per_item_full/SonxOSG.parquet +0 -0
  54. argus_uq/_data/per_item_full/SonxSP.parquet +0 -0
  55. argus_uq/_data/per_item_full/SonxUIV.parquet +0 -0
  56. argus_uq/_data/per_item_full/SonxV2.parquet +0 -0
  57. argus_uq/_data/per_item_full/UIxOSG.parquet +0 -0
  58. argus_uq/_data/per_item_full/UIxSP.parquet +0 -0
  59. argus_uq/_data/per_item_full/UIxUIV.parquet +0 -0
  60. argus_uq/_data/per_item_full/UIxV2.parquet +0 -0
  61. argus_uq/_data/subsets/osg_564.json +576 -0
  62. argus_uq/_data/subsets/ssp_300.json +368 -0
  63. argus_uq/_data/subsets/ssv2_300.json +360 -0
  64. argus_uq/_data/subsets/uiv_501.json +511 -0
  65. argus_uq/_version.py +3 -0
  66. argus_uq/analysis/__init__.py +3 -0
  67. argus_uq/api.py +338 -0
  68. argus_uq/cell.py +89 -0
  69. argus_uq/cli/__init__.py +9 -0
  70. argus_uq/cli/capture.py +87 -0
  71. argus_uq/cli/normalize.py +67 -0
  72. argus_uq/cli/prepare_data.py +307 -0
  73. argus_uq/cli/reproduce.py +147 -0
  74. argus_uq/cli/score.py +71 -0
  75. argus_uq/configs/__init__.py +1 -0
  76. argus_uq/conformal/__init__.py +86 -0
  77. argus_uq/conformal/base.py +82 -0
  78. argus_uq/conformal/crc.py +130 -0
  79. argus_uq/conformal/disk_cqr.py +114 -0
  80. argus_uq/conformal/disk_fixed.py +79 -0
  81. argus_uq/conformal/disk_normalized.py +92 -0
  82. argus_uq/data/__init__.py +98 -0
  83. argus_uq/data/cells.py +137 -0
  84. argus_uq/data/derived.py +75 -0
  85. argus_uq/data/per_item.py +53 -0
  86. argus_uq/envs/__init__.py +28 -0
  87. argus_uq/envs/base.py +138 -0
  88. argus_uq/envs/osworld_g.py +84 -0
  89. argus_uq/envs/screenspot_pro.py +74 -0
  90. argus_uq/envs/screenspot_v2.py +98 -0
  91. argus_uq/envs/ui_vision.py +77 -0
  92. argus_uq/estimators/__init__.py +62 -0
  93. argus_uq/estimators/attention.py +243 -0
  94. argus_uq/estimators/base.py +119 -0
  95. argus_uq/estimators/consistency.py +164 -0
  96. argus_uq/estimators/density.py +385 -0
  97. argus_uq/estimators/logit.py +124 -0
  98. argus_uq/estimators/prompt.py +96 -0
  99. argus_uq/estimators/sampling.py +226 -0
  100. argus_uq/estimators/vlm_native.py +154 -0
  101. argus_uq/factory.py +65 -0
  102. argus_uq/manager.py +184 -0
  103. argus_uq/models/__init__.py +51 -0
  104. argus_uq/models/base.py +125 -0
  105. argus_uq/models/closed_source/__init__.py +7 -0
  106. argus_uq/models/closed_source/anthropic.py +84 -0
  107. argus_uq/models/closed_source/base.py +156 -0
  108. argus_uq/models/closed_source/google.py +84 -0
  109. argus_uq/models/closed_source/openai.py +100 -0
  110. argus_uq/models/os_atlas.py +20 -0
  111. argus_uq/models/points_gui.py +20 -0
  112. argus_uq/models/qwen25_vl.py +35 -0
  113. argus_uq/models/ui_tars.py +18 -0
  114. argus_uq/models/whitebox.py +154 -0
  115. argus_uq/records.py +127 -0
  116. argus_uq/scoring/__init__.py +24 -0
  117. argus_uq/scoring/cached.py +115 -0
  118. argus_uq/scoring/multi_seed.py +84 -0
  119. argus_uq/stat_calculators/__init__.py +14 -0
  120. argus_uq/stat_calculators/base.py +84 -0
  121. argus_uq/ue_metrics/__init__.py +96 -0
  122. argus_uq/ue_metrics/_helpers.py +62 -0
  123. argus_uq/ue_metrics/aurc.py +37 -0
  124. argus_uq/ue_metrics/auroc.py +50 -0
  125. argus_uq/ue_metrics/ause.py +67 -0
  126. argus_uq/ue_metrics/bootstrap.py +35 -0
  127. argus_uq/ue_metrics/brier.py +24 -0
  128. argus_uq/ue_metrics/ece.py +37 -0
  129. argus_uq/ue_metrics/isotonic.py +156 -0
  130. argus_uq/ue_metrics/prr.py +77 -0
  131. argus_uq/ue_metrics/selective.py +29 -0
  132. argus_uq/ue_metrics/spearman.py +24 -0
  133. argus_uq/utils/__init__.py +21 -0
  134. argus_uq/utils/paths.py +63 -0
  135. argus_uq-0.1.0.dist-info/METADATA +233 -0
  136. argus_uq-0.1.0.dist-info/RECORD +142 -0
  137. argus_uq-0.1.0.dist-info/WHEEL +5 -0
  138. argus_uq-0.1.0.dist-info/entry_points.txt +5 -0
  139. argus_uq-0.1.0.dist-info/licenses/LICENSE +21 -0
  140. argus_uq-0.1.0.dist-info/licenses/LICENSE-DATA +19 -0
  141. argus_uq-0.1.0.dist-info/licenses/NOTICES.md +71 -0
  142. argus_uq-0.1.0.dist-info/top_level.txt +1 -0
argus_uq/__init__.py ADDED
@@ -0,0 +1,24 @@
1
+ """argus-uq: uncertainty quantification benchmark for VLM GUI grounding agents.
2
+
3
+ This is the public API surface:
4
+
5
+ >>> import argus_uq
6
+ >>> cell = argus_uq.load("PTxOSG")
7
+ >>> result = argus_uq.score(cell, method="saplma", seeds=50)
8
+ >>> disk = argus_uq.conformal_disk(cell, plug_in="saplma", alpha=0.10,
9
+ ... variant="normalized")
10
+
11
+ Anything else is internal and may change between versions.
12
+ """
13
+
14
+ from argus_uq._version import __version__
15
+ from argus_uq.api import conformal_disk, load, score
16
+ from argus_uq.cell import Cell
17
+
18
+ __all__ = [
19
+ "__version__",
20
+ "load",
21
+ "score",
22
+ "conformal_disk",
23
+ "Cell",
24
+ ]
@@ -0,0 +1,17 @@
1
+ cell_id,model,dataset,auroc_top1_method,auroc_top1_family,auroc_top1_mean,ause_top1_method,ause_top1_family,ause_top1_mean,agree
2
+ PTxOSG,POINTS,OSWorld-G,mahalanobis_rmd,density,0.8201373037736674,mahalanobis_rmd,density,1.4487720944651619,True
3
+ PTxSP,POINTS,ScreenSpot-Pro,saplma,density,0.8808854744870688,saplma,density,1.6838910919779315,True
4
+ PTxV2,POINTS,ScreenSpot-v2,sep,density,0.8433300497022345,saplma,density,1.332322370584984,False
5
+ PxUIV,POINTS,UI-Vision-EG,saplma,density,0.7955309159924083,lexical_similarity,sampling,1.4168912611244562,False
6
+ Q72xOSG,Qwen72B,OSWorld-G,sep,density,0.8376607634789452,semantic_entropy_weighted,sampling,1.749988588908934,False
7
+ Q72xSP,Qwen72B,ScreenSpot-Pro,saplma,density,0.8891665717143327,focus,attention,1.407749804900822,False
8
+ Q72xV2,Qwen72B,ScreenSpot-v2,saplma,density,0.7637849289967934,hedge,vlm-native,0.8890315900395884,False
9
+ Q72xUIV,Qwen72B,UI-Vision-EG,saplma,density,0.8344653392330383,ccp,hybrid,1.3430051881673335,False
10
+ Q7xOSG,Qwen7B,OSWorld-G,sep,density,0.7863550899811979,ccp,hybrid,1.402473220902292,False
11
+ Q7xSP,Qwen7B,ScreenSpot-Pro,sep,density,0.8827668215648495,cocoa,hybrid,1.0375822654416558,False
12
+ Q7xV2,Qwen7B,ScreenSpot-v2,saplma,density,0.8172319440025454,ccp,hybrid,0.9560727530561476,False
13
+ Q7xUIV,Qwen7B,UI-Vision-EG,mean_token_entropy,logit,0.8162140790742527,cocoa,hybrid,1.066601426198668,False
14
+ UIxOSG,UI-TARS,OSWorld-G,sep,density,0.8631467889908256,saplma,density,1.416380223362568,False
15
+ UIxSP,UI-TARS,ScreenSpot-Pro,saplma,density,0.8767605188067445,cocoa_1mca,hybrid,1.0618054809004414,False
16
+ UIxV2,UI-TARS,ScreenSpot-v2,cocoa_1mca,hybrid,0.8415201703110341,ccp,hybrid,0.8938736198442734,False
17
+ UIxUIV,UI-TARS,UI-Vision-EG,cocoa,hybrid,0.8247519379844961,semantic_entropy_weighted,sampling,0.9240945459830204,False
@@ -0,0 +1,13 @@
1
+ cell_id,vendor,dataset,auroc_top1_method,auroc_top1_family,auroc_top1_mean,ause_top1_method,ause_top1_family,ause_top1_mean,agree
2
+ anthropic_osg,anthropic,OSWorld-G,ccp,hybrid,0.6160367762128325,ccp,hybrid,0.5412948484672749,True
3
+ anthropic_sp,anthropic,ScreenSpot-Pro,ccp,hybrid,0.7311190250079139,ccp,hybrid,0.27250981315425593,True
4
+ anthropic_uiv,anthropic,UI-Vision-EG,ccp,hybrid,0.6514536767087787,ccp,hybrid,0.3927690638283077,True
5
+ anthropic_v2,anthropic,ScreenSpot-v2,verbalized_2s,verbalised,0.6475272727272727,verbalized_2s,verbalised,0.39333125761636145,True
6
+ gemini_osg,gemini,OSWorld-G,verbalized_1s,verbalised,0.8797956891317547,verbalized_1s,verbalised,0.31201317376375803,True
7
+ gemini_sp,gemini,ScreenSpot-Pro,verbalized_2s,verbalised,0.8563269027737787,verbalized_2s,verbalised,0.26330898218945825,True
8
+ gemini_uiv,gemini,UI-Vision-EG,ccp,hybrid,0.7912628865979382,ccp,hybrid,0.38706408864293623,True
9
+ gemini_v2_300,gemini,ScreenSpot-v2,verbalized_1s,verbalised,0.9662894248608535,verbalized_2s,verbalised,0.18429915760503326,False
10
+ openai_osg,openai,OSWorld-G,verbalized_2s,verbalised,0.7702987818383167,verbalized_2s,verbalised,0.6148893564188218,True
11
+ openai_sp_300,openai,ScreenSpot-Pro,ccp,hybrid,0.7349424342105263,ccp,hybrid,0.26681712953825637,True
12
+ openai_uiv,openai,UI-Vision-EG,verbalized_1s,verbalised,0.7225125112601342,lexical_similarity,sampling,0.4627543299999304,False
13
+ openai_v2_300,openai,ScreenSpot-v2,ccp,hybrid,0.797570612748881,lexical_similarity,sampling,0.3478509370405503,False