relevant_pages string | evidence_pages string | score int64 | doc_id string | doc_type string | question string | answer string | evidence_sources string | final_answer string | turn2_draft_answer string | turn1_colqwen_query null | turn1_colqwen_retrieval_results null | turn1_llm_query_input null | turn1_llm_retrieval_results string | turn1_llm_raw_output string | turn1_memory_out string | turn2_memory_in string | turn2_vlm_prompt_input string | turn2_vlm_raw_output string | turn2_final_answer string | turn2_response_type string | turn2_updated_question string | turn2_notes string | turn2_vlm_turn1_input_image0_ref string | turn2_vlm_turn1_input_image10_ref string | turn2_vlm_turn1_input_image11_ref string | turn2_vlm_turn1_input_image12_ref string | turn2_vlm_turn1_input_image13_ref string | turn2_vlm_turn1_input_image14_ref string | turn2_vlm_turn1_input_image15_ref string | turn2_vlm_turn1_input_image16_ref string | turn2_vlm_turn1_input_image17_ref string | turn2_vlm_turn1_input_image18_ref string | turn2_vlm_turn1_input_image19_ref string | turn2_vlm_turn1_input_image1_ref string | turn2_vlm_turn1_input_image20_ref string | turn2_vlm_turn1_input_image21_ref string | turn2_vlm_turn1_input_image22_ref string | turn2_vlm_turn1_input_image23_ref string | turn2_vlm_turn1_input_image2_ref string | turn2_vlm_turn1_input_image3_ref string | turn2_vlm_turn1_input_image4_ref string | turn2_vlm_turn1_input_image5_ref string | turn2_vlm_turn1_input_image6_ref string | turn2_vlm_turn1_input_image7_ref string | turn2_vlm_turn1_input_image8_ref string | turn2_vlm_turn1_input_image9_ref string | turn2_vlm_turn1_input_messages string | turn2_vlm_turn1_prompt string | turn2_vlm_turn1_raw_output string |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
[5] | [5] | 1 | PH_2016.06.08_Economy-Final.pdf | Research report / Introduction | According to the report, how do 5% of the Latinos see economic upward mobility for their children? | Less well-off | ["Chart"] | According to the report, 5% of the Latinos see their children being less well off financially than they themselves are now. | According to the report, 5% of the Latinos see their children being less well off financially than they themselves are now. | null | null | null | {"relevant_pages": [5], "document_summary": "", "source": "retrieval_gt_use"} | [retrieval_gt_use] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
According to the report, how do 5% of the Latinos see economic upward mobility... | <scratchpad>
1. Key elements observed:
- Pie chart showing % of Latinos who expect their children to be better off financially.
- Text mentioning 72% of Latinos expect their children to be better off financially.
- Note explaining the survey methodology and sample size.
- Document summary indicating the foc... | According to the report, 5% of the Latinos see their children being less well off financially than they themselves are now. | answer | null | null | [PH_2016.06.08_Economy-Final_page5.png] | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | [] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
According to the report, how do 5% of the Latinos see economic upward mobility... | <scratchpad>
1. Key elements observed:
- Pie chart showing % of Latinos who expect their children to be better off financially.
- Text mentioning 72% of Latinos expect their children to be better off financially.
- Note explaining the survey methodology and sample size.
- Document summary indicating the foc... | ||
[19, 20] | [19, 20] | 0 | PH_2016.06.08_Economy-Final.pdf | Research report / Introduction | According to the report, which one is greater in population in the survey? Foreign born Latinos, or the Latinos interviewed by cellphone? | Latinos interviewed by cellphone | ["Table"] | {"doc_id": "PH_2016.06.08_Economy-Final.pdf", "question": "According to the report, which one is greater in population in the survey? Foreign born Latinos, or the Latinos interviewed by cellphone?", "final_answer": null, "response_type": "query_update", "vlm_raw_output": "<scratchpad>\n1. Key elements observed in the p... | {"doc_id": "PH_2016.06.08_Economy-Final.pdf", "question": "According to the report, which one is greater in population in the survey? Foreign born Latinos, or the Latinos interviewed by cellphone?", "final_answer": null, "response_type": "query_update", "vlm_raw_output": "<scratchpad>\n1. Key elements observed in the p... | null | null | null | {"relevant_pages": [19, 20], "document_summary": "", "source": "retrieval_gt_use"} | [retrieval_gt_use] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
According to the report, which one is greater in population in the survey? For... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- Page 19 contains a table titled "Interviews by Strata" showing the breakdown of interviews by landline and cellphone categories.
- Page 20 contains a table titled "2015 National Survey of Latinos details" showing the sample si... | null | query_update | [Provide a rewritten long query that PRESERVES THE ORIGINAL MEANING of the question but adds specific details or keywords to help retrieve new relevant pages. The information retrieved from this new query must directly answer the original question.] | [IF using query_update, provide concise notes about what you've learned so far, what information is still missing, and your reasoning for the updated query. These notes will be appended to the document summary in the next iteration to maintain context across searches.] | [PH_2016.06.08_Economy-Final_page19.png] | null | null | null | null | null | null | null | null | null | null | [PH_2016.06.08_Economy-Final_page20.png] | null | null | null | null | null | null | null | null | null | null | null | null | [] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
According to the report, which one is greater in population in the survey? For... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- Page 19 contains a table titled "Interviews by Strata" showing the breakdown of interviews by landline and cellphone categories.
- Page 20 contains a table titled "2015 National Survey of Latinos details" showing the sample si... | ||
[14] | [14] | 0 | PH_2016.06.08_Economy-Final.pdf | Research report / Introduction | From this report, which subgroup among Hispanics has gained most confidence from 2008 to 2015? | Some college or more | ["Chart"] | The subgroup that has gained the most confidence from 2008 to 2015 is the 2nd generation, with a 14 percentage point increase from 67% in 2008 to 86% in 2015. | The subgroup that has gained the most confidence from 2008 to 2015 is the 2nd generation, with a 14 percentage point increase from 67% in 2008 to 86% in 2015. | null | null | null | {"relevant_pages": [14], "document_summary": "", "source": "retrieval_gt_use"} | [retrieval_gt_use] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
From this report, which subgroup among Hispanics has gained most confidence fr... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- A bar chart showing the percentage of Latinos who expect their family's financial situation to improve "a lot" or "some" in the next year, comparing 2008 and 2015.
- Specific data points for different subgroups of Hispanics, s... | The subgroup that has gained the most confidence from 2008 to 2015 is the 2nd generation, with a 14 percentage point increase from 67% in 2008 to 86% in 2015. | answer | null | null | [PH_2016.06.08_Economy-Final_page14.png] | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | [] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
From this report, which subgroup among Hispanics has gained most confidence fr... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- A bar chart showing the percentage of Latinos who expect their family's financial situation to improve "a lot" or "some" in the next year, comparing 2008 and 2015.
- Specific data points for different subgroups of Hispanics, s... | ||
[2, 22, 23] | [2, 22, 23] | 0 | PH_2016.06.08_Economy-Final.pdf | Research report / Introduction | Among all 12 references in this report, how many are from its own research center? | 8 | ["Pure-text (Plain-text)"] | The provided pages do not contain enough information to determine the total number of references from the Pew Research Center among all 12 references. We need to check the remaining pages to find out if there are any more references from the Pew Research Center. | The provided pages do not contain enough information to determine the total number of references from the Pew Research Center among all 12 references. We need to check the remaining pages to find out if there are any more references from the Pew Research Center. | null | null | null | {"relevant_pages": [2, 22, 23], "document_summary": "", "source": "retrieval_gt_use"} | [retrieval_gt_use] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
Among all 12 references in this report, how many are from its own research cen... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- Page 21 contains the "Appendix A: References" section.
- Page 22 contains additional references.
- The references list includes various sources, some from the Pew Research Center.
2. Specific details related to the questio... | The provided pages do not contain enough information to determine the total number of references from the Pew Research Center among all 12 references. We need to check the remaining pages to find out if there are any more references from the Pew Research Center. | answer | null | null | [PH_2016.06.08_Economy-Final_page2.png] | null | null | null | null | null | null | null | null | null | null | [PH_2016.06.08_Economy-Final_page22.png] | null | null | null | null | [PH_2016.06.08_Economy-Final_page23.png] | null | null | null | null | null | null | null | [] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
Among all 12 references in this report, how many are from its own research cen... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- Page 21 contains the "Appendix A: References" section.
- Page 22 contains additional references.
- The references list includes various sources, some from the Pew Research Center.
2. Specific details related to the questio... | ||
[4, 7, 9, 10, 13, 14] | [4, 7, 9, 10, 13, 14] | 0 | PH_2016.06.08_Economy-Final.pdf | Research report / Introduction | Among all charts in this report, how many compare between the general public with the Latinos/Hispanics? A few small charts sharing the same chart title is considered one chart. | 6 | ["Chart", "Generalized-text (Layout)"] | There are two charts comparing the general public with the Latinos/Hispanics in the document. One chart is on Page 4, titled "Hispanics more confident in personal finances than in 2008, see better economic times ahead," and another chart is on Page 12, titled "Financial expectations rising faster among Hispanics since ... | There are two charts comparing the general public with the Latinos/Hispanics in the document. One chart is on Page 4, titled "Hispanics more confident in personal finances than in 2008, see better economic times ahead," and another chart is on Page 12, titled "Financial expectations rising faster among Hispanics since ... | null | null | null | {"relevant_pages": [4, 7, 9, 10, 13, 14], "document_summary": "", "source": "retrieval_gt_use"} | [retrieval_gt_use] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
Among all charts in this report, how many compare between the general public w... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- Page 4: Comparison of personal financial situation between the general public and Hispanics.
- Page 7: Comparison of economic conditions between Hispanics and whites.
- Page 8: Comparison of Latino perceptions of personal f... | There are two charts comparing the general public with the Latinos/Hispanics in the document. One chart is on Page 4, titled "Hispanics more confident in personal finances than in 2008, see better economic times ahead," and another chart is on Page 12, titled "Financial expectations rising faster among Hispanics since ... | answer | null | null | [PH_2016.06.08_Economy-Final_page4.png] | null | null | null | null | null | null | null | null | null | null | [PH_2016.06.08_Economy-Final_page7.png] | null | null | null | null | [PH_2016.06.08_Economy-Final_page9.png] | [PH_2016.06.08_Economy-Final_page10.png] | [PH_2016.06.08_Economy-Final_page13.png] | [PH_2016.06.08_Economy-Final_page14.png] | null | null | null | null | [] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
Among all charts in this report, how many compare between the general public w... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- Page 4: Comparison of personal financial situation between the general public and Hispanics.
- Page 7: Comparison of economic conditions between Hispanics and whites.
- Page 8: Comparison of Latino perceptions of personal f... | ||
[16] | [16] | 1 | PH_2016.06.08_Economy-Final.pdf | Research report / Introduction | According to this report, among the Hispanics shaped by four current personal financial situation, which situation involves the highest percentage that expects their future financial situation to get a lot worse? | Poor Financial Condition | ["Chart"] | The situation involving the highest percentage that expects their future financial situation to get a lot worse is "Poor financial condition," with 8% of Hispanics in this category expecting their financial situation to get a lot worse in the next year. | The situation involving the highest percentage that expects their future financial situation to get a lot worse is "Poor financial condition," with 8% of Hispanics in this category expecting their financial situation to get a lot worse in the next year. | null | null | null | {"relevant_pages": [16], "document_summary": "", "source": "retrieval_gt_use"} | [retrieval_gt_use] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
According to this report, among the Hispanics shaped by four current personal ... | <scratchpad>
1. Key elements observed: The document discusses future financial expectations among Hispanics based on their current personal financial situation. A bar chart is present showing percentages of Hispanics who expect their family's financial situation to improve or worsen in the next year, categorized by the... | The situation involving the highest percentage that expects their future financial situation to get a lot worse is "Poor financial condition," with 8% of Hispanics in this category expecting their financial situation to get a lot worse in the next year. | answer | null | null | [PH_2016.06.08_Economy-Final_page16.png] | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | [] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
According to this report, among the Hispanics shaped by four current personal ... | <scratchpad>
1. Key elements observed: The document discusses future financial expectations among Hispanics based on their current personal financial situation. A bar chart is present showing percentages of Hispanics who expect their family's financial situation to improve or worsen in the next year, categorized by the... | ||
[11] | [11] | 0 | PH_2016.06.08_Economy-Final.pdf | Research report / Introduction | According to this report, from 2014 to 2015, one group has the most significant drop of percentage of households claiming their income was falling behind cost of living.
Which group is this, and what is the percentage droped? | ['White', '10%'] | ["Chart"] | The Hispanic population experienced no significant change in the percentage of households claiming their income was falling behind the cost of living between 2014 and 2015. The percentage remained at 53% in both years. | The Hispanic population experienced no significant change in the percentage of households claiming their income was falling behind the cost of living between 2014 and 2015. The percentage remained at 53% in both years. | null | null | null | {"relevant_pages": [11], "document_summary": "", "source": "retrieval_gt_use"} | [retrieval_gt_use] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
According to this report, from 2014 to 2015, one group has the most significan... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- A chart comparing the percentage of households claiming their income was falling behind the cost of living in 2014 and 2015 for Hispanic, White, and Black populations.
- Text discussing the findings of the surveys conducted by... | The Hispanic population experienced no significant change in the percentage of households claiming their income was falling behind the cost of living between 2014 and 2015. The percentage remained at 53% in both years. | answer | null | null | [PH_2016.06.08_Economy-Final_page11.png] | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | [] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
According to this report, from 2014 to 2015, one group has the most significan... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- A chart comparing the percentage of households claiming their income was falling behind the cost of living in 2014 and 2015 for Hispanic, White, and Black populations.
- Text discussing the findings of the surveys conducted by... | ||
[3, 5] | [3, 5] | 0 | Independents-Report.pdf | Research report / Introduction | What's the percentage of people who are democrats and voted in the last election compared to the entire population in 2018? | 18.29% | ["Pure-text (Plain-text)"] | The document does not provide the total population in 2018 or the percentage of Democrats who voted in the last election. Therefore, it is not possible to calculate the percentage of people who are democrats and voted in the last election compared to the entire population in 2018. | The document does not provide the total population in 2018 or the percentage of Democrats who voted in the last election. Therefore, it is not possible to calculate the percentage of people who are democrats and voted in the last election compared to the entire population in 2018. | null | null | null | {"relevant_pages": [3, 5], "document_summary": "", "source": "retrieval_gt_use"} | [retrieval_gt_use] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
What's the percentage of people who are democrats and voted in the last electi... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- Page 3 discusses the political leanings of Americans in 2018.
- Page 5 provides data on voter registration and voting among partisan leaners and independents.
- The chart on page 3 shows the distribution of political leanin... | The document does not provide the total population in 2018 or the percentage of Democrats who voted in the last election. Therefore, it is not possible to calculate the percentage of people who are democrats and voted in the last election compared to the entire population in 2018. | answer | null | null | [Independents-Report_page3.png] | null | null | null | null | null | null | null | null | null | null | [Independents-Report_page5.png] | null | null | null | null | null | null | null | null | null | null | null | null | [] | You are an AI assistant capable of analyzing documents and extracting relevant information to answer questions. You will be provided with document pages and a question about these pages.
Consider this question about the document:
<question>
What's the percentage of people who are democrats and voted in the last electi... | <scratchpad>
1. Key elements observed in the provided page images and extracted text:
- Page 3 discusses the political leanings of Americans in 2018.
- Page 5 provides data on voter registration and voting among partisan leaners and independents.
- The chart on page 3 shows the distribution of political leanin... | ||
[] | [] | 1 | Independents-Report.pdf | Research report / Introduction | "What's the percentage of people who are democrats and voted in the last election compared to the en(...TRUNCATED) | Not answerable | [] | "{\"doc_id\": \"Independents-Report.pdf\", \"question\": \"What's the percentage of people who are d(...TRUNCATED) | "{\"doc_id\": \"Independents-Report.pdf\", \"question\": \"What's the percentage of people who are d(...TRUNCATED) | null | null | null | {"relevant_pages": [], "document_summary": "", "source": "retrieval_gt_use"} | [retrieval_gt_use] | "You are an AI assistant capable of analyzing documents and extracting relevant information to answe(...TRUNCATED) | "<scratchpad>\n1. The provided page images and extracted text do not contain any information related(...TRUNCATED) | null | query_update | "[Provide a rewritten long query that PRESERVES THE ORIGINAL MEANING of the question but adds specif(...TRUNCATED) | "[IF using query_update, provide concise notes about what you've learned so far, what information is(...TRUNCATED) | [no_image] | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | [] | "You are an AI assistant capable of analyzing documents and extracting relevant information to answe(...TRUNCATED) | "<scratchpad>\n1. The provided page images and extracted text do not contain any information related(...TRUNCATED) | ||
[] | [] | 1 | Independents-Report.pdf | Research report / Introduction | "From this report, among Clinton, G.W.Bush, and Obama, whose support rate by the democrats increased(...TRUNCATED) | Not answerable | [] | "{\"doc_id\": \"Independents-Report.pdf\", \"question\": \"From this report, among Clinton, G.W.Bush(...TRUNCATED) | "{\"doc_id\": \"Independents-Report.pdf\", \"question\": \"From this report, among Clinton, G.W.Bush(...TRUNCATED) | null | null | null | {"relevant_pages": [], "document_summary": "", "source": "retrieval_gt_use"} | [retrieval_gt_use] | "You are an AI assistant capable of analyzing documents and extracting relevant information to answe(...TRUNCATED) | "<scratchpad>\n1. The provided page text and images do not contain any specific data or tables relat(...TRUNCATED) | null | query_update | "[Provide a rewritten long query that PRESERVES THE ORIGINAL MEANING of the question but adds specif(...TRUNCATED) | "[IF using query_update, provide concise notes about what you've learned so far, what information is(...TRUNCATED) | [no_image] | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | [] | "You are an AI assistant capable of analyzing documents and extracting relevant information to answe(...TRUNCATED) | "<scratchpad>\n1. The provided page text and images do not contain any specific data or tables relat(...TRUNCATED) |
End of preview. Expand in Data Studio
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
MMLongBench – 2025-12-20 11:32 UTC
Average accuracy: 51.12% (1072 samples with scores)
Subset metrics by evidence source:
Pure-text (Plain-text): samples=302, accuracy=46.36%
Figure: samples=299, accuracy=39.13%
Table: samples=217, accuracy=35.48%
Chart: samples=175, accuracy=33.14%
Generalized-text (Layout): samples=119, accuracy=39.50%
Subset metrics by evidence pages length:
no_pages: samples=226, accuracy=83.19%
single_page: samples=489, accuracy=53.99%
multiple_pages: samples=357, accuracy=26.89%
Done: Results saved to /hub_data2/seohyun/outputs/var1_with_gt/simpledoc_eval/MMLongBench/eval_results.jsonl
Results source: /hub_data2/seohyun/outputs/var1_with_gt/results.json
dataset_info: features: - name: relevant_pages dtype: string - name: evidence_pages dtype: string - name: score dtype: int64 - name: doc_id dtype: string - name: doc_type dtype: string - name: question dtype: string - name: answer dtype: string - name: evidence_sources dtype: string - name: final_answer dtype: string - name: turn2_draft_answer dtype: string - name: turn1_colqwen_query dtype: 'null' - name: turn1_colqwen_retrieval_results dtype: 'null' - name: turn1_llm_query_input dtype: 'null' - name: turn1_llm_retrieval_results dtype: string - name: turn1_llm_raw_output dtype: string - name: turn1_memory_out dtype: string - name: turn2_memory_in dtype: string - name: turn2_vlm_prompt_input dtype: string - name: turn2_vlm_raw_output dtype: string - name: turn2_final_answer dtype: string - name: turn2_response_type dtype: string - name: turn2_updated_question dtype: string - name: turn2_notes dtype: string - name: turn2_vlm_turn1_input_image0_ref dtype: string - name: turn2_vlm_turn1_input_image10_ref dtype: string - name: turn2_vlm_turn1_input_image11_ref dtype: string - name: turn2_vlm_turn1_input_image12_ref dtype: string - name: turn2_vlm_turn1_input_image13_ref dtype: string - name: turn2_vlm_turn1_input_image14_ref dtype: string - name: turn2_vlm_turn1_input_image15_ref dtype: string - name: turn2_vlm_turn1_input_image16_ref dtype: string - name: turn2_vlm_turn1_input_image17_ref dtype: string - name: turn2_vlm_turn1_input_image18_ref dtype: string - name: turn2_vlm_turn1_input_image19_ref dtype: string - name: turn2_vlm_turn1_input_image1_ref dtype: string - name: turn2_vlm_turn1_input_image20_ref dtype: string - name: turn2_vlm_turn1_input_image21_ref dtype: string - name: turn2_vlm_turn1_input_image22_ref dtype: string - name: turn2_vlm_turn1_input_image23_ref dtype: string - name: turn2_vlm_turn1_input_image2_ref dtype: string - name: turn2_vlm_turn1_input_image3_ref dtype: string - name: turn2_vlm_turn1_input_image4_ref dtype: string - name: turn2_vlm_turn1_input_image5_ref dtype: string - name: turn2_vlm_turn1_input_image6_ref dtype: string - name: turn2_vlm_turn1_input_image7_ref dtype: string - name: turn2_vlm_turn1_input_image8_ref dtype: string - name: turn2_vlm_turn1_input_image9_ref dtype: string - name: turn2_vlm_turn1_input_messages dtype: string - name: turn2_vlm_turn1_prompt dtype: string - name: turn2_vlm_turn1_raw_output dtype: string splits: - name: train num_bytes: 22150180 num_examples: 1073 download_size: 3291842 dataset_size: 22150180 configs: - config_name: default data_files: - split: train path: data/train-*
- Downloads last month
- 15