WoW
A Chinese educational corpus dataset built on an educational knowledge graph as its classification skeleton. It covers multiple subjects across all school stages, supports multimodal signals, and is designed for full-lifecycle training of foundation models.
Timeline (last 90 days)
The WoW (World Knowledge) dataset is a Chinese educational corpus dataset proposed in a paper from RSS.
The core construction idea of the WoW dataset is to use an educational knowledge graph as the classification skeleton and systematically collect and complete the missing educational modules for each school stage.
The WoW dataset is designed for the full-cycle training of foundation models, covering the complete needs from pre-training to subsequent stages.
In the construction of WoW, the educational knowledge graph serves as a gap detection tool, locating missing modules by comparing graph nodes with the coverage of existing corpora.
Compared with the previous version, the new version of WoW is enhanced in all three dimensions: systematicity, quality and scale.
The new version of WoW introduces multimodal signals and is no longer limited to plain text.
WoW explicitly supports every stage of model development.
The paper's experimental results show that WoW significantly improves model performance on diverse benchmarks.
The paper abstract does not disclose the specific benchmark names or the magnitude of performance improvement in the WoW experiments.
The authors provide an interactive platform that supports knowledge-graph-based data selection and retrieval.
2 more timeline events
All Facts (12)
The paper abstract does not elaborate on the construction details of the graph, the specific types of multimodal signals, or the practical usability of the platform.
50%UnverifiedThe core construction idea of the WoW dataset is to use an educational knowledge graph as the classification skeleton and systematically collect and complete the missing educational modules for each school stage.
50%UnverifiedThe WoW dataset is designed for the full-cycle training of foundation models, covering the complete needs from pre-training to subsequent stages.
50%UnverifiedIn the construction of WoW, the educational knowledge graph serves as a gap detection tool, locating missing modules by comparing graph nodes with the coverage of existing corpora.
50%UnverifiedCompared with the previous version, the new version of WoW is enhanced in all three dimensions: systematicity, quality and scale.
50%UnverifiedThe new version of WoW introduces multimodal signals and is no longer limited to plain text.
50%UnverifiedWoW explicitly supports every stage of model development.
50%UnverifiedThe paper's experimental results show that WoW significantly improves model performance on diverse benchmarks.
50%UnverifiedThe paper abstract does not disclose the specific benchmark names or the magnitude of performance improvement in the WoW experiments.
50%UnverifiedThe WoW (World Knowledge) dataset is a Chinese educational corpus dataset proposed in a paper from RSS.
50%UnverifiedThe core function of the interactive platform is to associate corpora with graph nodes, so that users can quickly locate and filter the corresponding corpora by clicking on subject, school stage or knowledge point nodes.
50%UnverifiedThe authors provide an interactive platform that supports knowledge-graph-based data selection and retrieval.
50%