trainingDataSource
P21073
predicate
Indicates the origin or provider from which the training data for a model or system is obtained.
All labels observed (2)
| Label | Occurrences |
|---|---|
| trainingDataSource canonical | 14 |
| trainingSignalSource | 1 |
Description generation (PDg)
The one-sentence description above was generated by prompting gpt-5.1 with the predicate name and this instruction.
Instruction
Given a predicate that represents a relationship or action between entities, generate a one-sentence description explaining its meaning. # Instructions Focus on describing the relationship, not the entities themselves. # Response Format Begin the description with \' Indicates...\'
Input
Predicate: trainingDataSource
Generated description
Indicates the origin or provider from which the training data for a model or system is obtained.
Sample triples (15)
| Subject | Object |
|---|---|
| GPT-2 | WebText dataset ⓘ |
| GPT-3 | Common Crawl ⓘ |
| GPT-3 | WebText-like corpora ⓘ |
| GPT-3 | books ⓘ |
| GPT-3 | Wikipedia ⓘ |
| AlphaZero | self‑generated game data ⓘ |
| RoBERTa | BookCorpus NERFINISHED ⓘ |
| RoBERTa | English Wikipedia NERFINISHED ⓘ |
| RoBERTa | CC-News ⓘ |
| RoBERTa | OpenWebText NERFINISHED ⓘ |
| RoBERTa | Stories corpus ⓘ |
| OPT | publicly available text data ⓘ |
| OPT | licensed text data ⓘ |
| Bloom | ROOTS corpus NERFINISHED ⓘ |
| Contrastive Predictive Coding | data itself via predicate surface "trainingSignalSource" ⓘ |