Dataset Viewer
Auto-converted to Parquet Duplicate
data_source
stringclasses
1 value
prompt
stringlengths
988
128k
ability
stringclasses
1 value
reward_model
dict
extra_info
dict
dbpedia50
"\n You are given a directed graph as two CSV-like sections in this order:\n\n 1) Node table ((...TRUNCATED)
graph_path
{ "style": "rule" }
{"entities":["1000_Smiling_Knuckles","Anglo-Burmese_people"],"ground_truth":[["1000_Smiling_Knuckles(...TRUNCATED)
dbpedia50
"\n You are given a directed graph as two CSV-like sections in this order:\n\n 1) Node table ((...TRUNCATED)
graph_path
{ "style": "rule" }
{"entities":["100_(30_Rock)","Frank_Westheimer"],"ground_truth":[["100_(30_Rock)","guest","Rachel_Dr(...TRUNCATED)
dbpedia50
"\n You are given a directed graph as two CSV-like sections in this order:\n\n 1) Node table ((...TRUNCATED)
graph_path
{ "style": "rule" }
{"entities":["12_Monkeys","The_Killer_That_Stalked_New_York"],"ground_truth":[["12_Monkeys","distrib(...TRUNCATED)
dbpedia50
"\n You are given a directed graph as two CSV-like sections in this order:\n\n 1) Node table ((...TRUNCATED)
graph_path
{ "style": "rule" }
{"entities":["17_Days_(song)","Vanity_6"],"ground_truth":[["17_Days_(song)","writer","Doctor_Fink"],(...TRUNCATED)
dbpedia50
"\n You are given a directed graph as two CSV-like sections in this order:\n\n 1) Node table ((...TRUNCATED)
graph_path
{ "style": "rule" }
{"entities":["17th_Combat_Service_Support_Brigade_(Australia)","BBC"],"ground_truth":[["17th_Combat_(...TRUNCATED)
dbpedia50
"\n You are given a directed graph as two CSV-like sections in this order:\n\n 1) Node table ((...TRUNCATED)
graph_path
{ "style": "rule" }
{"entities":["1932_Ford","Modena"],"ground_truth":[["1932_Ford","bodyStyle","Convertible"],["Convert(...TRUNCATED)
dbpedia50
"\n You are given a directed graph as two CSV-like sections in this order:\n\n 1) Node table ((...TRUNCATED)
graph_path
{ "style": "rule" }
{"entities":["1968_United_States_Grand_Prix","1976_Dutch_Grand_Prix"],"ground_truth":[["1968_United_(...TRUNCATED)
dbpedia50
"\n You are given a directed graph as two CSV-like sections in this order:\n\n 1) Node table ((...TRUNCATED)
graph_path
{ "style": "rule" }
{"entities":["1968_United_States_Grand_Prix","Mario_Andretti"],"ground_truth":[["1968_United_States_(...TRUNCATED)
dbpedia50
"\n You are given a directed graph as two CSV-like sections in this order:\n\n 1) Node table ((...TRUNCATED)
graph_path
{ "style": "rule" }
{"entities":["1973_Spanish_Grand_Prix","Roman_Hryhorchuk"],"ground_truth":[["1973_Spanish_Grand_Prix(...TRUNCATED)
dbpedia50
"\n You are given a directed graph as two CSV-like sections in this order:\n\n 1) Node table ((...TRUNCATED)
graph_path
{ "style": "rule" }
{"entities":["1976_Dutch_Grand_Prix","Mario_Andretti"],"ground_truth":[["1976_Dutch_Grand_Prix","thi(...TRUNCATED)
End of preview. Expand in Data Studio

DBpedia50 (GNN subgraph in prompt)

Same schema and non-graph fields as tangyx/DBpedia50_2entities. The Graph: section in prompt is rebuilt from the named subgraph in gnn/data/dbpedia50/{train,test}.json (ReaRev export), not the original HF demo graph.

Build locally:

python dbpedia50/export_hf_prompts_gnn_subgraph.py \
  --out_dir dbpedia50/parquet \
  --gnn_data_dir gnn/data/dbpedia50

extra_info["valid_edges"] is replaced with the same named triples as the new graph; other extra_info fields match the base dataset.

Downloads last month
36