ai-forever commited on
Commit
7c4d697
1 Parent(s): 79253b3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +11 -11
README.md CHANGED
@@ -617,35 +617,35 @@ The datasets are divided into subsets based on context lengths: 4k, 8k, 16k, 32k
617
  <img src="table.png" width="800" />
618
 
619
  ### Group I: Simple Information Retrieval
620
- - **Passkey**: Extract a relevant piece of code number from a long text fragment.
621
  - **PasskeyWithLibrusec**: Similar to Passkey but with added noise from Librusec texts.
622
 
623
  ### Group II: Question Answering and Multiple Choice
624
- - **MatreshkaNames**: Identify the person in dialogues based on the discussed topic.
625
  - **MatreshkaYesNo**: Indicate whether a specific topic was mentioned in the dialog.
626
  - **LibrusecHistory**: Answer questions based on historical texts.
627
- - **ruTREC**: Few-shot in-context learning for topic classification. Created by translating the TREC dataset from [LongBench](https://github.com/THUDM/LongBench).
628
- - **ruSciFi**: Answer true/false based on context and general world knowledge. Translation of SciFi dataset from [L-Eval](https://github.com/OpenLMLab/LEval).
629
  - **ruSciAbstractRetrieval**: Retrieve relevant paragraphs from scientific abstracts.
630
- - **ruTPO**: Multiple-choice questions similar to TOEFL exams. Translation of the TPO dataset from [L-Eval](https://github.com/OpenLMLab/LEval).
631
- - **ruQuALITY**: Multiple-choice QA tasks based on detailed texts. Created by translating the QuALITY dataset from [L-Eval](https://github.com/OpenLMLab/LEval).
632
 
633
  ### Group III: Multi-hop Question Answering
634
  - **ruBABILongQA**: 5 long-context reasoning tasks for QA using facts hidden among irrelevant information.
635
  - **LongContextMultiQ**: Multi-hop QA based on Wikidata and Wikipedia.
636
  - **LibrusecMHQA**: Multi-hop QA requiring information distributed across several text parts.
637
- - **ru2WikiMultihopQA**: Translation of the 2WikiMultihopQA dataset from [LongBench](https://github.com/THUDM/LongBench).
638
 
639
  ### Group IV: Complex Reasoning and Mathematical Problems
640
  - **ruSciPassageCount**: Count unique paragraphs in a long text.
641
- - **ruQasper**: Question Answering over academic research papers. Created by translating the Qasper dataset from [LongBench](https://github.com/THUDM/LongBench).
642
- - **ruGSM100**: Solve math problems using Chain-of-Thought reasoning.
643
 
644
 
645
 
646
  ## Usage
647
 
648
- The LIBRA benchmark is available under the MIT license. Researchers and developers can use these datasets to evaluate the long-context understanding abilities of various LLMs. The datasets, codebase, and public leaderboard are open-source to guide forthcoming research in this area.
649
 
650
  ## Citation
651
 
@@ -664,6 +664,6 @@ year={2024}
664
 
665
  The datasets are published under the MIT license.
666
 
667
- ## Acknowledgments
668
 
669
  For more details and code, please visit our [GitHub repository](https://github.com/ai-forever/LIBRA/).
 
617
  <img src="table.png" width="800" />
618
 
619
  ### Group I: Simple Information Retrieval
620
+ - **Passkey**: Extract a relevant piece of code number from a long text fragment. Based on the original [PassKey test](https://github.com/CStanKonrad/long_llama/blob/main/examples/passkey.py) from the m LongLLaMA’s GitHub repo.
621
  - **PasskeyWithLibrusec**: Similar to Passkey but with added noise from Librusec texts.
622
 
623
  ### Group II: Question Answering and Multiple Choice
624
+ - **MatreshkaNames**: Identify the person in dialogues based on the discussed topic. We used [Matreshka](https://huggingface.co/datasets/zjkarina/matreshka) dataset and [Russian Names](https://www.kaggle.com/datasets/rai220/russian-cyrillic-names-and-sex/data) dataset to create this and the next task.
625
  - **MatreshkaYesNo**: Indicate whether a specific topic was mentioned in the dialog.
626
  - **LibrusecHistory**: Answer questions based on historical texts.
627
+ - **ruTREC**: Few-shot in-context learning for topic classification. Created by translating the [TREC dataset](https://huggingface.co/datasets/THUDM/LongBench/viewer/trec_e) from LongBench.
628
+ - **ruSciFi**: Answer true/false based on context and general world knowledge. Translation of [SciFi dataset](https://huggingface.co/datasets/L4NLP/LEval/viewer/sci_f) from L-Eval which originally was based on [SF-Gram](https://github.com/nschaetti/SFGram-dataset).
629
  - **ruSciAbstractRetrieval**: Retrieve relevant paragraphs from scientific abstracts.
630
+ - **ruTPO**: Multiple-choice questions similar to TOEFL exams. Translation of the [TPO dataset](https://huggingface.co/datasets/L4NLP/LEval/viewer/tpo) from L-Eval.
631
+ - **ruQuALITY**: Multiple-choice QA tasks based on detailed texts. Created by translating the [QuALITY dataset](https://huggingface.co/datasets/L4NLP/LEval/viewer/quality) from L-Eval.
632
 
633
  ### Group III: Multi-hop Question Answering
634
  - **ruBABILongQA**: 5 long-context reasoning tasks for QA using facts hidden among irrelevant information.
635
  - **LongContextMultiQ**: Multi-hop QA based on Wikidata and Wikipedia.
636
  - **LibrusecMHQA**: Multi-hop QA requiring information distributed across several text parts.
637
+ - **ru2WikiMultihopQA**: Translation of the [2WikiMultihopQA dataset](https://huggingface.co/datasets/THUDM/LongBench/viewer/2wikimqa_e) from LongBench.
638
 
639
  ### Group IV: Complex Reasoning and Mathematical Problems
640
  - **ruSciPassageCount**: Count unique paragraphs in a long text.
641
+ - **ruQasper**: Question Answering over academic research papers. Created by translating the [Qasper dataset](https://huggingface.co/datasets/THUDM/LongBench/viewer/qasper_e) from LongBench.
642
+ - **ruGSM100**: Solve math problems using Chain-of-Thought reasoning. Created by translating the [GSM100](https://huggingface.co/datasets/L4NLP/LEval/viewer/gsm100) dataset from L-Eval.
643
 
644
 
645
 
646
  ## Usage
647
 
648
+ Researchers and developers can use these datasets to evaluate the long-context understanding abilities of various LLMs. The datasets, codebase, and public leaderboard are open-source to guide forthcoming research in this area.
649
 
650
  ## Citation
651
 
 
664
 
665
  The datasets are published under the MIT license.
666
 
667
+ ## GitHub
668
 
669
  For more details and code, please visit our [GitHub repository](https://github.com/ai-forever/LIBRA/).