List your company

Factual Consistency Evaluation on TRUE benchmark

98.4PAWS (AUC-ROC)ALIGNSCORE-large
45.037658.891372.74586.5987May 26, 2023Jul 9, 2023Aug 23, 2023Oct 7, 2023Nov 21, 2023Jan 5, 2024Feb 19, 2024

Evaluation results

MethodLinks
ALIGNSCORE-largeType=Ours2023.0598.482.978.698.394.992.185.176.189.583.582.587.483.8———————
ALIGNSCORE-baseType=Ours2023.0597.380.876.197.894.69083.179.987.779.682.486.382.5———————
M + A + WLM size=T5 XXL2024.0289.95—76.4588.55———79.49——83.5685.69——90.687.2386.7379.4485.5695.07
SummaC-ZSType=NLI2023.058977.681.897.292.886.987.1587675.383.282.278.2———————
SummaC-CONVType=NLI2023.0588.279.177.597.5928981.267.277.77681.682.578.7———————
WANLILM size=T5 XXL2024.0287.29—84.8887.38———73.29——82.4783.72——88.5972.1882.8574.6192.494.97
GNLILM size=T5 XXL2024.0287.1—77.487.69———79.97——8485.75——91.3885.4887.0379.2888.5795.34
ANLILM size=T5 XXL2024.0286.35—74.6890.59———76.41——79.7782.98——87.982.0884.6875.7981.0693.46
QAFactEvalType=QA2023.0586.180.975.873.68688.581.867.383.976.18180.179.4———————
M + A + WLM size=T5 LARGE2024.0285.86—79.8186.48———78.28——81.1382.2——83.5772.2882.2772.6187.2594.63
ANLILM size=T5 LARGE2024.0284.72—78.4389.63———75.49——78.2181.46——81.7874.6981.8171.685.6394.03
GNLILM size=T5 LARGE2024.0284.63—79.7785.86———79.49——83.4584.72——90.1481.3384.0279.7588.7694.73
WANLILM size=T5 LARGE2024.0283.17—82.1682.79———67.7——80.3779.72——80.3174.4671.1172.8689.1593.82
MNLILM size=T5 XXL2024.0283.11—76.8484.58———78.31——81.7682.79——88.1879.0383.0772.3588.3295.13
MNLIType=NLI2023.0581.344.671.880.293.157.276.559.142.650.181.567.160.4———————
UniEvalType=Misc, Variant=Consistency2023.0580.181.270.479.192.188.180.466.886.576.773.679.578———————
BERTScoreType=Similarity Matching2023.0578.672.370.258.254.28468.652.570.644.386.467.268.6———————
BARTScoreType=Misc, Checkpoint=BART-Large-CNN2023.0577.178.965.164.266.187.860.863.583.960.286.772.273.4———————
ROUGE-LType=Similarity Matching2023.0575.980.460.659.756.483.665.462.877.659.38569.771.8———————
MNLILM size=T5 LARGE2024.0275.82—77.3281.6———79.27——82.1377.84——79.1558.1379.5661.5987.6593.97
BLEUType=Similarity Matching2023.0571.374.855.256.151.784.161.256.777.454.774.665.267.3———————
SimCSEType=Similarity Matching2023.0569.270.266.263.872.772.970.664.674.956.586.169.870.3———————
QuestEvalType=QA2023.056969.772.266.672.58477.264.864.555.283.970.971.4———————
ROUGE-2Type=Similarity Matching2023.0568.679.461.459.955.584.567.76578.460.282.869.472.4———————
BLEURTType=Regression2023.0568.46872.961.859.581.67365.571.256.286.669.571.9———————
CTCType=Misc, Variant=D-CNNDM2023.0563.179.866.86572.587.163.76577.367.77270.972.4———————
M + A + WLM size=T5 SMALL2024.0259.15—6672.71———69.69——79.0865.33——50.247.3561.7646.7576.0189.94
GNLILM size=T5 SMALL2024.0257.33—72.176.11———76.91——82.2172.06——67.3260.2272.3956.2981.2390.54
BLANCType=Misc2023.055663.362.955.753.682.163.854.260.950.973.761.664———————
DAEType=NLI2023.0555.860.357.760.277.877.954.78156.967.569.465.465.7———————
FactCCType=Misc2023.0553.468.659.354.758.770.75556.170.164.457.660.862.7———————
NER-OverlapType=Similarity Matching2023.0551.756.659.157.862.465.562.768.448.463.650.658.859.3———————
MNLILM size=T5 SMALL2024.0251.68—68.470.1———70.3——80.7763.48——49.6237.7658.345.9776.189.33
WANLILM size=T5 SMALL2024.0251.51—68.9767.85———69.9——65.7965.21——57.9954.1470.2148.9877.6284.35
ROUGE-1Type=Similarity Matching2023.0550.280.459.760.957.883.665.364.877.360.184.667.772———————
FEQAType=QA2023.055049.553.249.951.16350.548.850.149.45351.752.2———————
ANLILM size=T5 SMALL2024.0247.09—45.0253.5———53.99——57.6651.77——50.9554.6444.751.3455.3955.24
AllParameters=350M, Model calls=15, MC Dropout=true2023.05—————————————83.2——————
All (-MC)Parameters=350M, Model calls=1, MC Dropout=false2023.05—————————————82.7——————
Q2Parameters=930M (220+355+355), Model calls=#Q * (QI + 2)2023.05—————————————81.4——————
SummacZSParameters=355M, Model calls=#snt x #snt2023.05—————————————80.7——————
T5 ANLIParameters=11,000M, Model calls=12023.05—————————————81.5——————