Added Contamination Evidence on MMLU of ChatGPT/GPT4 from "Investigating data contamination in modern benchmarks for large language models"

.gitignore

@@ -1,2 +1,3 @@
 *.pyc
-*.json
+*.json
+*.lock

app.py

@@ -22,11 +22,11 @@ def filter_dataframe(dataframe, eval_dataset, cont_source, checkboxes):
     """
     if isinstance(eval_dataset, str):
         dataframe = dataframe[
-            dataframe["Evaluation Dataset"].str.contains(eval_dataset)
+            dataframe["Evaluation Dataset"].str.contains(f"(?i){eval_dataset}")
         ]
     if isinstance(cont_source, str):
         dataframe = dataframe[
-            dataframe["Contaminated Source"].str.contains(cont_source)
+            dataframe["Contaminated Source"].str.contains(f"(?i){cont_source}")
         ]
     if isinstance(checkboxes, list) and "Exclude model-based evidences" in checkboxes:
         dataframe = dataframe[dataframe["Approach"] != "model-based"]

contamination_report.csv

@@ -1,5 +1,22 @@
 Evaluation Dataset;Subset;Contaminated Source;Model or corpus;Train Split;Development Split;Test Split;Approach;Reference;PR
 
+gsm8k;;GPT-4;model;79.00;;;model-based;https://arxiv.org/abs/2311.06233;8
+ucinlp/drop;;GPT-4;model;;44.00;;model-based;https://arxiv.org/abs/2311.06233;8
+openai_humaneval;;GPT-4;model;;;56.71;model-based;https://arxiv.org/abs/2311.06233;8
+imdb;;GPT-4;model;;;82.00;model-based;https://arxiv.org/abs/2311.06233;8
+imdb;;GPT-3.5;model;;;55.00;model-based;https://arxiv.org/abs/2311.06233;8
+ag_news;;GPT-4;model;;;91.00;model-based;https://arxiv.org/abs/2311.06233;8
+ag_news;;GPT-3.5;model;;;82.00;model-based;https://arxiv.org/abs/2311.06233;8
+yelp_review_full;;GPT-4;model;;;80.00;model-based;https://arxiv.org/abs/2311.06233;8
+yelp_review_full;;GPT-3.5;model;;;13.00;model-based;https://arxiv.org/abs/2311.06233;8
+nyu-mll/glue;rte;GPT-4;model;;60.00;;model-based;https://arxiv.org/abs/2311.06233;8
+nyu-mll/glue;rte;GPT-3.5;model;;71.00;;model-based;https://arxiv.org/abs/2311.06233;8
+nyu-mll/glue;wnli;GPT-4;model;;50.70;;model-based;https://arxiv.org/abs/2311.06233;8
+nyu-mll/glue;wnli;GPT-3.5;model;;12.68;;model-based;https://arxiv.org/abs/2311.06233;8
+samsum;;GPT-4;model;;;77.00;model-based;https://arxiv.org/abs/2311.06233;8
+samsum;;GPT-3.5;model;;;74.00;model-based;https://arxiv.org/abs/2311.06233;8
+EdinburghNLP/xsum;;GPT-4;model;;;95.00;model-based;https://arxiv.org/abs/2311.06233;8
+EdinburghNLP/xsum;;GPT-3.5;model;;;79.00;model-based;https://arxiv.org/abs/2311.06233;8
 
 allenai/ai2_arc;;CommonCrawl;corpus;;;28.7;data-based;https://arxiv.org/abs/2310.17589;5
 tau/commonsense_qa;;CommonCrawl;corpus;;1.6;;data-based;https://arxiv.org/abs/2310.17589;5
@@ -436,30 +453,171 @@ zest;;EleutherAI/pile;corpus;;;0.0;data-based;https://arxiv.org/abs/2310.20707;2
 zest;;togethercomputer/RedPajama-Data-V2;corpus;;;0.0;data-based;https://arxiv.org/abs/2310.20707;2
  
  
-imdb;;GPT-4;model;100.0;;0.0;model-based;https://arxiv.org/pdf/2308.08493;3
-imdb;;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/pdf/2308.08493;3
+imdb;;GPT-4;model;100.0;;0.0;model-based;https://arxiv.org/abs/2308.08493;3
+imdb;;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/abs/2308.08493;3
  
-ag_news;;GPT-4;model;100.0;;100.0;model-based;https://arxiv.org/pdf/2308.08493;3
-ag_news;;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/pdf/2308.08493;3
+ag_news;;GPT-4;model;100.0;;100.0;model-based;https://arxiv.org/abs/2308.08493;3
+ag_news;;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/abs/2308.08493;3
  
-yelp_review_full;;GPT-4;model;0.0;;0.0;model-based;https://arxiv.org/pdf/2308.08493;3
-yelp_review_full;;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/pdf/2308.08493;3
+yelp_review_full;;GPT-4;model;0.0;;0.0;model-based;https://arxiv.org/abs/2308.08493;3
+yelp_review_full;;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/abs/2308.08493;3
  
-nyu-mll/glue;rte;GPT-4;model;100.0;;0.0;model-based;https://arxiv.org/pdf/2308.08493;3
-nyu-mll/glue;rte;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/pdf/2308.08493;3
+nyu-mll/glue;rte;GPT-4;model;100.0;;0.0;model-based;https://arxiv.org/abs/2308.08493;3
+nyu-mll/glue;rte;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/abs/2308.08493;3
  
-nyu-mll/glue;wnli;GPT-4;model;100.0;;100.0;model-based;https://arxiv.org/pdf/2308.08493;3
-nyu-mll/glue;wnli;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/pdf/2308.08493;3
+nyu-mll/glue;wnli;GPT-4;model;100.0;;100.0;model-based;https://arxiv.org/abs/2308.08493;3
+nyu-mll/glue;wnli;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/abs/2308.08493;3
  
-samsum;;GPT-4;model;0.0;;0.0;model-based;https://arxiv.org/pdf/2308.08493;3
-samsum;;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/pdf/2308.08493;3
+samsum;;GPT-4;model;0.0;;0.0;model-based;https://arxiv.org/abs/2308.08493;3
+samsum;;GPT-3.5;model;0.0;;0.0;model-based;https://arxiv.org/abs/2308.08493;3
  
-EdinburghNLP/xsum;;GPT-4;model;0.0;;100.0;model-based;https://arxiv.org/pdf/2308.08493;3
-EdinburghNLP/xsum;;GPT-3.5;model;0.0;;100.0;model-based;https://arxiv.org/pdf/2308.08493;3
+EdinburghNLP/xsum;;GPT-4;model;0.0;;100.0;model-based;https://arxiv.org/abs/2308.08493;3
+EdinburghNLP/xsum;;GPT-3.5;model;0.0;;100.0;model-based;https://arxiv.org/abs/2308.08493;3
 
-bigbio/mednli;;GPT-4;model;0.0;0.0;0.0;model-based;https://arxiv.org/pdf/2308.08493;8
-bigbio/mednli;;GPT-3.5;model;0.0;0.0;0.0;model-based;https://arxiv.org/pdf/2308.08493;8
+bigbio/mednli;;GPT-4;model;0.0;0.0;0.0;model-based;https://arxiv.org/abs/2308.08493;8
+bigbio/mednli;;GPT-3.5;model;0.0;0.0;0.0;model-based;https://arxiv.org/abs/2308.08493;8
 
 RadNLI;;GPT-4;model;0.0;0.0;0.0;model-based;https://arxiv.org/pdf/2308.08493;8
 RadNLI;;GPT-3.5;model;0.0;0.0;0.0;model-based;https://arxiv.org/pdf/2308.08493;8
- 
+
+
+cais/mmlu;;GPT-3.5;model;;;52.0;model-based;https://arxiv.org/abs/2311.09783;10
+winogrande;;GPT-3.5;model;;;9.0;model-based;https://arxiv.org/abs/2311.09783;10
+truthful_qa;;GPT-3.5;model;;;12.0;model-based;https://arxiv.org/abs/2311.09783;10
+allenai/openbookqa;;GPT-3.5;model;;;1.0;model-based;https://arxiv.org/abs/2311.09783;10
+
+cais/mmlu;;GPT-4;model;;;57.0;model-based;https://arxiv.org/abs/2311.09783;10
+truthful_qa;;GPT-4;model;;;10.0;model-based;https://arxiv.org/abs/2311.09783;10
+winogrande;;GPT-4;model;;;12.0;model-based;https://arxiv.org/abs/2311.09783;10
+allenai/openbookqa;;GPT-4;model;;;1.0;model-based;https://arxiv.org/abs/2311.09783;10
+Rowan/hellaswag;;GPT-4;model;;;2.0;model-based;https://arxiv.org/abs/2311.09783;10
+
+
+allenai/openbookqa;;LLaMa 2-13B;model;;;4.0;model-based;https://arxiv.org/abs/2311.09783;10
+truthful_qa;;LLaMa 2-13B;model;;;2.0;model-based;https://arxiv.org/abs/2311.09783;10
+winogrande;;LLaMa 2-13B;model;;;1.0;model-based;https://arxiv.org/abs/2311.09783;10
+
+truthful_qa;;Mistral-7B;model;;;15.0;model-based;https://arxiv.org/abs/2311.09783;10
+allenai/openbookqa;;Mistral-7B;model;;;10.0;model-based;https://arxiv.org/abs/2311.09783;10
+winogrande;;Mistral-7B;model;;;3.0;model-based;https://arxiv.org/abs/2311.09783;10
+cais/mmlu;;Mistral-7B;model;;;1.0;model-based;https://arxiv.org/abs/2311.09783;10
+RadNLI;;GPT-4;model;0.0;0.0;0.0;model-based;https://arxiv.org/abs/2308.08493;8
+RadNLI;;GPT-3.5;model;0.0;0.0;0.0;model-based;https://arxiv.org/abs/2308.08493;8
+RadNLI;;GPT-4;model;0.0;0.0;0.0;model-based;https://arxiv.org/abs/2308.08493;8
+RadNLI;;GPT-3.5;model;0.0;0.0;0.0;model-based;https://arxiv.org/abs/2308.08493;8
+
+
+openai_humaneval;;EleutherAI/pile;corpus;;;12.2;data-based;https://arxiv.org/abs/2403.04811;12
+mbpp;;EleutherAI/pile;corpus;;;3.6;data-based;https://arxiv.org/abs/2403.04811;12
+openai_humaneval;;bigcode/the-stack;corpus;;;18.9;data-based;https://arxiv.org/abs/2403.04811;12
+mbpp;;bigcode/the-stack;corpus;;;20.8;data-based;https://arxiv.org/abs/2403.04811;12
+
+quac;;GPT-3;model;;99.0;;data-based;https://arxiv.org/abs/2005.14165;13
+rajpurkar/squad_v2;;GPT-3;model;;94.0;;data-based;https://arxiv.org/abs/2005.14165;13
+ucinlp/drop;;GPT-3;model;;93.0;;data-based;https://arxiv.org/abs/2005.14165;13
+Symbol Insertion;;GPT-3;model;;86.0;;data-based;https://arxiv.org/abs/2005.14165;13
+stanfordnlp/coqa;;GPT-3;model;;64.0;;data-based;https://arxiv.org/abs/2005.14165;13
+super_glue;record;GPT-3;model;;61.0;;data-based;https://arxiv.org/abs/2005.14165;13
+winograd_wsc;;GPT-3;model;;;60.0;data-based;https://arxiv.org/abs/2005.14165;13
+super_glue;boolq;GPT-3;model;;60.0;;data-based;https://arxiv.org/abs/2005.14165;13
+super_glue;multirc;GPT-3;model;;59.0;;data-based;https://arxiv.org/abs/2005.14165;13
+race;high;GPT-3;model;;;45.0;data-based;https://arxiv.org/abs/2005.14165;13
+cimec/lambada;;GPT-3;model;;;43.0;data-based;https://arxiv.org/abs/2005.14165;13
+super_glue;wsc;GPT-3;model;;40.0;;data-based;https://arxiv.org/abs/2005.14165;13
+piqa;;GPT-3;model;;29.0;;data-based;https://arxiv.org/abs/2005.14165;13
+wmt/wmt16;en-de;GPT-3;model;;;25.0;data-based;https://arxiv.org/abs/2005.14165;13
+wmt/wmt16;de-en;GPT-3;model;;;25.0;data-based;https://arxiv.org/abs/2005.14165;13
+race;middle;GPT-3;model;;;25.0;data-based;https://arxiv.org/abs/2005.14165;13
+rmanluo/RoG-webqsp;;GPT-3;model;;;21.0;data-based;https://arxiv.org/abs/2005.14165;13
+wmt/wmt16;en-ro;GPT-3;model;;;21.0;data-based;https://arxiv.org/abs/2005.14165;13
+wmt/wmt16;ro-en;GPT-3;model;;;21.0;data-based;https://arxiv.org/abs/2005.14165;13
+facebook/anli;test_r1;GPT-3;model;;;20.0;data-based;https://arxiv.org/abs/2005.14165;13
+facebook/anli;test_r2;GPT-3;model;;;18.0;data-based;https://arxiv.org/abs/2005.14165;13
+mandarjoshi/trivia_qa;;GPT-3;model;;17.0;;data-based;https://arxiv.org/abs/2005.14165;13
+facebook/anli;test_r3;GPT-3;model;;;16.0;data-based;https://arxiv.org/abs/2005.14165;13
+wmt/wmt16;fr-en;GPT-3;model;;;14.0;data-based;https://arxiv.org/abs/2005.14165;13
+wmt/wmt16;en-fr;GPT-3;model;;;14.0;data-based;https://arxiv.org/abs/2005.14165;13
+super_glue;rte;GPT-3;model;;8.0;;data-based;https://arxiv.org/abs/2005.14165;13
+super_glue;wic;GPT-3;model;;8.0;;data-based;https://arxiv.org/abs/2005.14165;13
+super_glue;cb;GPT-3;model;;7.0;;data-based;https://arxiv.org/abs/2005.14165;13
+Reversed Words;;GPT-3;model;;7.0;;data-based;https://arxiv.org/abs/2005.14165;13
+Anagrams 2;;GPT-3;model;;7.0;;data-based;https://arxiv.org/abs/2005.14165;13
+allenai/openbookqa;;GPT-3;model;;;6.0;data-based;https://arxiv.org/abs/2005.14165;13
+ibragim-bad/arc_easy;;GPT-3;model;;;4.0;data-based;https://arxiv.org/abs/2005.14165;13
+Anagrams 1;;GPT-3;model;;3.0;;data-based;https://arxiv.org/abs/2005.14165;13
+ibragim-bad/arc_challenge;;GPT-3;model;;;3.0;data-based;https://arxiv.org/abs/2005.14165;13
+super_glue;copa;GPT-3;model;;3.0;;data-based;https://arxiv.org/abs/2005.14165;13
+Rowan/hellaswag;;GPT-3;model;;2.0;;data-based;https://arxiv.org/abs/2005.14165;13
+natural_questions;;GPT-3;model;;;1.0;data-based;https://arxiv.org/abs/2005.14165;13
+Cycled Letters;;GPT-3;model;;1.0;;data-based;https://arxiv.org/abs/2005.14165;13
+SAT Analogies;;GPT-3;model;;1.0;;data-based;https://arxiv.org/abs/2005.14165;13
+
+EdinburghNLP/xsum;;PaLM 2;model;;;42.0;data-based;https://arxiv.org/abs/2305.10403;13
+csebuetnlp/xlsum;;PaLM 2;model;;;46.9;data-based;https://arxiv.org/abs/2305.10403;13
+wiki_lingua;;PaLM 2;model;;;9.0;data-based;https://arxiv.org/abs/2305.10403;13
+
+winograd_wsc;;PaLM;model;;;38.5;data-based;https://arxiv.org/abs/2204.02311;13
+rmanluo/RoG-webqsp;;PaLM;model;;;26.7;data-based;https://arxiv.org/abs/2204.02311;13
+super_glue;wsc;PaLM;model;;;36.8;data-based;https://arxiv.org/abs/2204.02311;13
+mandarjoshi/trivia_qa;;PaLM;model;;19.9;;data-based;https://arxiv.org/abs/2204.02311;13
+rajpurkar/squad_v2;;PaLM;model;;85.2;;data-based;https://arxiv.org/abs/2204.02311;13
+super_glue;record;PaLM;model;;43.4;;data-based;https://arxiv.org/abs/2204.02311;13
+cimec/lambada;;PaLM;model;;;29.3;data-based;https://arxiv.org/abs/2204.02311;13
+super_glue;cb;PaLM;model;;48.2;;data-based;https://arxiv.org/abs/2204.02311;13
+ibragim-bad/arc_easy;;PaLM;model;;;30.4;data-based;https://arxiv.org/abs/2204.02311;13
+ibragim-bad/arc_challenge;;PaLM;model;;;24.7;data-based;https://arxiv.org/abs/2204.02311;13
+
+winograd_wsc;;GLaM;model;;67.3;;data-based;https://arxiv.org/abs/2112.06905;13
+winogrande;;GLaM;model;;;0.3;data-based;https://arxiv.org/abs/2112.06905;13
+super_glue;wic;GLaM;model;;8.2;;data-based;https://arxiv.org/abs/2112.06905;13
+super_glue;wsc;GLaM;model;;57.5;;data-based;https://arxiv.org/abs/2112.06905;13
+mandarjoshi/trivia_qa;;GLaM;model;;18.8;;data-based;https://arxiv.org/abs/2112.06905;13
+story_cloze;;GLaM;model;;100.0;;data-based;https://arxiv.org/abs/2112.06905;13
+rajpurkar/squad_v2;;GLaM;model;;94.6;;data-based;https://arxiv.org/abs/2112.06905;13
+super_glue;record;GLaM;model;;98.6;;data-based;https://arxiv.org/abs/2112.06905;13
+super_glue;rte;GLaM;model;;54.9;;data-based;https://arxiv.org/abs/2112.06905;13
+race;middle;GLaM;model;;58.4;;data-based;https://arxiv.org/abs/2112.06905;13
+race;high;GLaM;model;;74.0;;data-based;https://arxiv.org/abs/2112.06905;13
+quac;;GLaM;model;;99.9;;data-based;https://arxiv.org/abs/2112.06905;13
+piqa;;GLaM;model;;49.8;;data-based;https://arxiv.org/abs/2112.06905;13
+allenai/openbookqa;;GLaM;model;;20.0;;data-based;https://arxiv.org/abs/2112.06905;13
+natural_questions;;GLaM;model;;3.9;;data-based;https://arxiv.org/abs/2112.06905;13
+super_glue;multirc;GLaM;model;;68.8;;data-based;https://arxiv.org/abs/2112.06905;13
+cimec/lambada;;GLaM;model;;;21.8;data-based;https://arxiv.org/abs/2112.06905;13
+Rowan/hellaswag;;GLaM;model;;19.8;;data-based;https://arxiv.org/abs/2112.06905;13
+stanfordnlp/coqa;;GLaM;model;;;75.0;data-based;https://arxiv.org/abs/2112.06905;13
+super_glue;copa;GLaM;model;;3.0;;data-based;https://arxiv.org/abs/2112.06905;13
+super_glue;cb;GLaM;model;;26.8;;data-based;https://arxiv.org/abs/2112.06905;13
+super_glue;boolq;GLaM;model;;92.1;;data-based;https://arxiv.org/abs/2112.06905;13
+ibragim-bad/arc_easy;;GLaM;model;;32.5;;data-based;https://arxiv.org/abs/2112.06905;13
+ibragim-bad/arc_challenge;;GLaM;model;;31.8;;data-based;https://arxiv.org/abs/2112.06905;13
+facebook/anli;dev_r3;GLaM;model;;40.7;;data-based;https://arxiv.org/abs/2112.06905;13
+facebook/anli;dev_r2;GLaM;model;;96.8;;data-based;https://arxiv.org/abs/2112.06905;13
+facebook/anli;dev_r1;GLaM;model;;96.2;;data-based;https://arxiv.org/abs/2112.06905;13
+
+winogrande;;FLAN;model;;;0.2;data-based;https://arxiv.org/abs/2109.01652;13
+mandarjoshi/trivia_qa;;FLAN;model;;22.8;;data-based;https://arxiv.org/abs/2109.01652;13
+story_cloze;;FLAN;model;;0.4;;data-based;https://arxiv.org/abs/2109.01652;13
+rajpurkar/squad_v2;;FLAN;model;;99.1;;data-based;https://arxiv.org/abs/2109.01652;13
+wmt/wmt16;ro-en;FLAN;model;;;12.4;data-based;https://arxiv.org/abs/2109.01652;13
+super_glue;record;FLAN;model;;68.0;;data-based;https://arxiv.org/abs/2109.01652;13
+super_glue;rte;FLAN;model;;33.9;;data-based;https://arxiv.org/abs/2109.01652;13
+piqa;;FLAN;model;;51.3;;data-based;https://arxiv.org/abs/2109.01652;13
+allenai/openbookqa;;FLAN;model;;15.0;;data-based;https://arxiv.org/abs/2109.01652;13
+natural_questions;;FLAN;model;;3.2;;data-based;https://arxiv.org/abs/2109.01652;13
+super_glue;multirc;FLAN;model;;59.3;;data-based;https://arxiv.org/abs/2109.01652;13
+Rowan/hellaswag;;FLAN;model;;34.5;;data-based;https://arxiv.org/abs/2109.01652;13
+wmt/wmt16;fr-en;FLAN;model;;;25.3;data-based;https://arxiv.org/abs/2109.01652;13
+wmt/wmt16;en-ro;FLAN;model;;;12.4;data-based;https://arxiv.org/abs/2109.01652;13
+wmt/wmt16;en-fr;FLAN;model;;;25.3;data-based;https://arxiv.org/abs/2109.01652;13
+wmt/wmt16;en-de;FLAN;model;;;14.3;data-based;https://arxiv.org/abs/2109.01652;13
+wmt/wmt16;de-en;FLAN;model;;;14.3;data-based;https://arxiv.org/abs/2109.01652;13
+ucinlp/drop;;FLAN;model;;99.4;;data-based;https://arxiv.org/abs/2109.01652;13
+super_glue;copa;FLAN;model;;9.0;;data-based;https://arxiv.org/abs/2109.01652;13
+super_glue;cb;FLAN;model;;5.4;;data-based;https://arxiv.org/abs/2109.01652;13
+super_glue;boolq;FLAN;model;;23.1;;data-based;https://arxiv.org/abs/2109.01652;13
+ibragim-bad/arc_easy;;FLAN;model;;20.2;;data-based;https://arxiv.org/abs/2109.01652;13
+ibragim-bad/arc_challenge;;FLAN;model;;15.6;;data-based;https://arxiv.org/abs/2109.01652;13
+facebook/anli;dev_r3;FLAN;model;;40.2;;data-based;https://arxiv.org/abs/2109.01652;13
+facebook/anli;dev_r2;FLAN;model;;97.9;;data-based;https://arxiv.org/abs/2109.01652;13
+facebook/anli;dev_r1;FLAN;model;;98.6;;data-based;https://arxiv.org/abs/2109.01652;13