Dataset Viewer
Auto-converted to Parquet Duplicate
prompt
stringlengths
3.58k
262k
chosen
listlengths
2
2
rejected
listlengths
2
2
chosen_model
stringclasses
6 values
rejected_model
stringclasses
5 values
id
stringlengths
36
36
"A user issued a query and a set of research papers were provided with salient content.\nThe user qu(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
sqa+claude37
llm-only_DeepSeek-V3-0324
dddca3a4-a844-5803-83fc-2af47dd8510c
"A user issued a query and a set of research papers were provided with salient content.\nThe user qu(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
sqa+claude37
llm-only_DeepSeek-V3-0324
26cf38ed-2cf1-516d-9101-0955d719ac96
"A user issued a query and a set of research papers were provided with salient content.\nThe user qu(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
llm-only_DeepSeek-V3-0324
sqa+claude37
0b06dfe2-6786-5414-8d24-29b1e148d7d7
"A user issued a query and a set of research papers were provided with salient content.\nThe user qu(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
sqa+claude37
llm-only_DeepSeek-V3-0324
d5123bb7-b9e5-519c-8397-036f03055db9
"A user issued a query and a set of research papers were provided with salient content.\nThe user qu(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
sqa+claude37
llm-only_DeepSeek-V3-0324
92061a64-6241-5ea9-ba22-3a8a9d0812d6
"A user issued a query and a set of research papers were provided with salient content.\nThe user qu(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
llm-only_DeepSeek-V3-0324
sqa+claude37
0e80883f-55d9-5f68-b35f-82b9b7251849
"A user issued a query and a set of research papers were provided with salient content.\nThe user qu(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
llm-only_DeepSeek-V3-0324
sqa+claude37
ac216bf3-3838-53b7-a11b-4e218f8b0db7
"A user issued a query and a set of research papers were provided with salient content.\nThe user qu(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
sqa+claude37
llm-only_DeepSeek-V3-0324
d86316a9-19a4-5b13-8323-2f3b79bcbed9
"A user issued a query and a set of research papers were provided with salient content.\nThe user qu(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
sqa+claude37
llm-only_DeepSeek-V3-0324
f6448394-8b82-5fed-be64-df17f4457339
"A user issued a query and a set of research papers were provided with salient content.\nThe user qu(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
[{"content":"A user issued a query and a set of research papers were provided with salient content.\(...TRUNCATED)
sqa+claude37
llm-only_DeepSeek-V3-0324
865f319d-a053-5bab-8f52-b69f2aea55ed
End of preview. Expand in Data Studio

AstaBrief DPO Data

This dataset contains DPO training data for AstaBrief, consisting of 6.6k real user query prompts and final cited reports.

User Query Curation

User queries included in this dataset are sourced from real queries issued to Ai2's OpenScholar and Asta ScholarQA systems until June 2025, and filtered in the same way as our SFT data.

We are left with a pool of 12k queries at the end of this filtering process and after setting 47k aside for SFT.

Report Generation

For each query in our set, we produce two reports. One report comes from the ScholarQA pipeline, typically backed by Claude 3.5 Sonnet or 3.7 Sonnet. The competing report is generated by feeding ScholarQA's retrieved literature excerpts to a different model: o3, o4-mini, DeepSeek-V3, or DeepSeek-R1, depending on the example.

Preference Generation

We generate preferences per report pair using two judge models with a general report comparison prompt – GPT-4.1 and DeepSeek-R1. Both models compare each pair of reports and pick a winner. We ensure that these LLM judges were align with human preferences (95% agreement) and only keep pairs where both judges agree, giving us a cleaner preference set containing 6.6k instances.

License

This dataset is licensed under CC BY-NC 4.0. It is intended for research and educational use in accordance with Ai2's Responsible Use Guidelines. The data contains synthetic outputs generated by third party models, which are subject to the terms and conditions of the respective model provider.

Downloads last month
35

Models trained or fine-tuned on allenai/AstaBrief_DPO_Mix

Collection including allenai/AstaBrief_DPO_Mix

Article mentioning allenai/AstaBrief_DPO_Mix