-
Notifications
You must be signed in to change notification settings - Fork 2
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
1 parent
4ef0a32
commit 0b8d801
Showing
24 changed files
with
2,080 additions
and
919 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,23 @@ | ||
# This file may be used to create an environment using: | ||
# $ conda create --name <env> --file <this file> | ||
# platform: linux-64 | ||
@EXPLICIT | ||
https://repo.anaconda.com/pkgs/main/linux-64/ld_impl_linux-64-2.33.1-h53a641e_7.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/_libgcc_mutex-0.1-main.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/ca-certificates-2021.7.5-h06a4308_1.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/libstdcxx-ng-9.1.0-hdf63c60_0.tar.bz2 | ||
https://repo.continuum.io/pkgs/main/linux-64/libgcc-ng-9.1.0-hdf63c60_0.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/libffi-3.2.1-hd88cf55_4.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/ncurses-6.2-he6710b0_1.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/openssl-1.0.2u-h7b6447c_0.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/xz-5.2.5-h7b6447c_0.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/zlib-1.2.11-h7b6447c_3.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/libedit-3.1.20181209-hc058e9b_0.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/readline-7.0-h7b6447c_5.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/tk-8.6.8-hbc83047_0.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/sqlite-3.31.1-h62c20be_1.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/python-3.6.3-h6c0c0dc_5.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/certifi-2021.5.30-py36h06a4308_0.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/noarch/wheel-0.35.1-py_0.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/setuptools-49.6.0-py36_0.tar.bz2 | ||
https://repo.anaconda.com/pkgs/main/linux-64/pip-20.2.2-py36_0.tar.bz2 |
20 changes: 12 additions & 8 deletions
20
double_faithfulness_metrics/multirc/contigious-dev-f1-metrics-description.json
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,38 +1,42 @@ | ||
{ | ||
"f1 macro avg - model labels": { | ||
"fixed-deeplift": 60.841, | ||
"fixed-lime": 55.433, | ||
"fixed-attention": 61.682, | ||
"fixed-ig": 59.283, | ||
"fixed-gradients": 57.084, | ||
"fixed-scaled attention": 59.496, | ||
"fixed-random": 60.938, | ||
"fixed-fixed-len_var-feat": 42.041, | ||
"fixed-fixed-len_var-feat_var-type": 34.468, | ||
"fixed-fixed-len_var-feat": 41.078, | ||
"fixed-fixed-len_var-feat_var-type": 33.462, | ||
"var-deeplift": 46.906, | ||
"var-lime": 41.413, | ||
"var-attention": 50.006, | ||
"var-ig": 45.583, | ||
"var-gradients": 45.18, | ||
"var-scaled attention": 47.564, | ||
"var-random": 48.382, | ||
"var-var-len_var-feat": 31.741, | ||
"var-var-len_var-feat_var-type": 27.098 | ||
"var-var-len_var-feat": 31.088, | ||
"var-var-len_var-feat_var-type": 26.5 | ||
}, | ||
"f1 macro avg - actual labels": { | ||
"fixed-deeplift": 56.789, | ||
"fixed-lime": 53.337, | ||
"fixed-attention": 57.202, | ||
"fixed-ig": 55.856, | ||
"fixed-gradients": 53.201, | ||
"fixed-scaled attention": 55.741, | ||
"fixed-random": 55.676, | ||
"fixed-fixed-len_var-feat": 46.263, | ||
"fixed-fixed-len_var-feat_var-type": 42.723, | ||
"fixed-fixed-len_var-feat": 45.804, | ||
"fixed-fixed-len_var-feat_var-type": 42.099, | ||
"var-deeplift": 51.116, | ||
"var-lime": 46.713, | ||
"var-attention": 52.893, | ||
"var-ig": 50.203, | ||
"var-gradients": 48.476, | ||
"var-scaled attention": 51.915, | ||
"var-random": 51.783, | ||
"var-var-len_var-feat": 43.023, | ||
"var-var-len_var-feat_var-type": 40.607 | ||
"var-var-len_var-feat": 42.948, | ||
"var-var-len_var-feat_var-type": 40.233 | ||
} | ||
} |
210 changes: 115 additions & 95 deletions
210
double_faithfulness_metrics/multirc/contigious-dev-faithfulness-metrics-description.json
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,162 +1,182 @@ | ||
{ | ||
"fixed-lime": { | ||
"sufficiency": { | ||
"mean": 0.11085775690785485, | ||
"std": 0.19380987336211145 | ||
"fixed-deeplift": { | ||
"comprehensiveness": { | ||
"mean": 0.5708503014596947, | ||
"std": 0.40622983297882476 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.12635902115296146, | ||
"std": 0.20305647933234272 | ||
} | ||
}, | ||
"fixed-lime": { | ||
"comprehensiveness": { | ||
"mean": 0.5863771542531726, | ||
"std": 0.4144303612791726 | ||
"mean": 0.5862558961329564, | ||
"std": 0.4143992123739367 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.11110520351084904, | ||
"std": 0.19389510120352713 | ||
} | ||
}, | ||
"fixed-attention": { | ||
"sufficiency": { | ||
"mean": 0.14935326796972878, | ||
"std": 0.23014501697197656 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.54822363556328, | ||
"std": 0.4054012030234662 | ||
"mean": 0.5481157222388134, | ||
"std": 0.4053917153418573 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.149572736610961, | ||
"std": 0.23021060780657382 | ||
} | ||
}, | ||
"fixed-ig": { | ||
"sufficiency": { | ||
"mean": 0.11152060733602513, | ||
"std": 0.19327101360991455 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.5830766712017098, | ||
"std": 0.4070952517192963 | ||
"mean": 0.5829574974832881, | ||
"std": 0.4070752921781835 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.11176561694157401, | ||
"std": 0.19336390089956032 | ||
} | ||
}, | ||
"fixed-gradients": { | ||
"sufficiency": { | ||
"mean": 0.11271092160973989, | ||
"std": 0.19455577996066112 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.5683918653592137, | ||
"std": 0.4144701058992363 | ||
"mean": 0.5682874672562472, | ||
"std": 0.41445382396748404 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.11300825692050266, | ||
"std": 0.1946987337861625 | ||
} | ||
}, | ||
"fixed-scaled attention": { | ||
"sufficiency": { | ||
"mean": 0.11853921326274494, | ||
"std": 0.20315889743785975 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.5511737595058142, | ||
"std": 0.41423814042591456 | ||
"mean": 0.5510775623105245, | ||
"std": 0.4142298296661397 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.11875628100944073, | ||
"std": 0.2032489100415157 | ||
} | ||
}, | ||
"fixed-random": { | ||
"sufficiency": { | ||
"mean": 0.1125270846427616, | ||
"std": 0.1958789389492079 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.5725935799325271, | ||
"std": 0.40806777343914896 | ||
"mean": 0.5724729113991164, | ||
"std": 0.40804452837153943 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.11277564690449358, | ||
"std": 0.1959607178687227 | ||
} | ||
}, | ||
"fixed-fixed-len_var-feat": { | ||
"sufficiency": { | ||
"mean": 0.14212839693064958, | ||
"std": 0.21408926738092865 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.6777123360380712, | ||
"std": 0.3831265888202201 | ||
"mean": 0.6828441882075921, | ||
"std": 0.3806118734168832 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.14770886702562278, | ||
"std": 0.21776328251523475 | ||
} | ||
}, | ||
"fixed-fixed-len_var-feat_var-type": { | ||
"sufficiency": { | ||
"mean": 0.16621332929362104, | ||
"std": 0.23099271289519063 | ||
"comprehensiveness": { | ||
"mean": 0.7384569092781771, | ||
"std": 0.3533037563683677 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.17647252350211667, | ||
"std": 0.23706898396815992 | ||
} | ||
}, | ||
"var-deeplift": { | ||
"comprehensiveness": { | ||
"mean": 0.7268847642651807, | ||
"std": 0.36077399020867795 | ||
"mean": 0.6626297399754025, | ||
"std": 0.39125732251714856 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.09831814960477779, | ||
"std": 0.1850340422130891 | ||
} | ||
}, | ||
"var-lime": { | ||
"sufficiency": { | ||
"mean": 0.0742763766680182, | ||
"std": 0.1658230492893937 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.658270898615268, | ||
"std": 0.40093258486123673 | ||
"mean": 0.6581062901951721, | ||
"std": 0.40093988352043936 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.07505030075786776, | ||
"std": 0.16600717812131904 | ||
} | ||
}, | ||
"var-attention": { | ||
"sufficiency": { | ||
"mean": 0.10674163969947464, | ||
"std": 0.19064829330521588 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.6181293375930368, | ||
"std": 0.3968848260321575 | ||
"mean": 0.6179474005087348, | ||
"std": 0.39690630380090086 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.10745532986051953, | ||
"std": 0.1908453526218175 | ||
} | ||
}, | ||
"var-ig": { | ||
"sufficiency": { | ||
"mean": 0.08675410257308236, | ||
"std": 0.17717385525329649 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.6708768093384752, | ||
"std": 0.3906119927114652 | ||
"mean": 0.6706880393183623, | ||
"std": 0.3906165246628001 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.08744349580951162, | ||
"std": 0.1772917683115485 | ||
} | ||
}, | ||
"var-gradients": { | ||
"sufficiency": { | ||
"mean": 0.08283040773414521, | ||
"std": 0.17211025615790934 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.6485212690978732, | ||
"std": 0.4026727710792673 | ||
"mean": 0.6483514554278175, | ||
"std": 0.4026750835518532 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.08363821395118706, | ||
"std": 0.17222866674338036 | ||
} | ||
}, | ||
"var-scaled attention": { | ||
"sufficiency": { | ||
"mean": 0.088175745282316, | ||
"std": 0.1781408343565686 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.639330701081299, | ||
"std": 0.39837204574386914 | ||
"mean": 0.6391521786915065, | ||
"std": 0.3983859724101448 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.08883505717095926, | ||
"std": 0.1783772329856403 | ||
} | ||
}, | ||
"var-random": { | ||
"sufficiency": { | ||
"mean": 0.08677869206029334, | ||
"std": 0.17376037597183477 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.6602227685384234, | ||
"std": 0.39202220232533375 | ||
"mean": 0.6600281645160797, | ||
"std": 0.39202528763826044 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.08748308008710975, | ||
"std": 0.1739491805021988 | ||
} | ||
}, | ||
"var-var-len_var-feat": { | ||
"sufficiency": { | ||
"mean": 0.09110130054647325, | ||
"std": 0.17759282684255423 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.7493078215482373, | ||
"std": 0.356383685796083 | ||
"mean": 0.7571034796139563, | ||
"std": 0.3519297311778094 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.09301501433294009, | ||
"std": 0.1808206715780162 | ||
} | ||
}, | ||
"var-var-len_var-feat_var-type": { | ||
"sufficiency": { | ||
"mean": 0.10328006265665282, | ||
"std": 0.18812752910940983 | ||
}, | ||
"comprehensiveness": { | ||
"mean": 0.7935850931773125, | ||
"std": 0.32531321101339833 | ||
"mean": 0.8022498643418685, | ||
"std": 0.3185238131956753 | ||
}, | ||
"sufficiency": { | ||
"mean": 0.10941112362793849, | ||
"std": 0.19222474951771257 | ||
} | ||
} | ||
} |
Oops, something went wrong.