MSMARCO-Passage Teacher Training IDs for Cross-Architecture Knowledge Distillation

Teacher train files containing query and passage IDs from the MSMARCO-Passage collection, created for the paper 'Improving Efficient Neural Ranking Models with Cross-Architecture Knowledge Distillation'. Sebastian Hofstätter and colleagues at TU Wien developed this resource to support research into distilling ranking knowledge between different neural architectures. The associated documentation and code are hosted on GitHub.

Use Cases

• Training student ranking models based on teacher scores derived from the provided IDs.
• Reproducing experiments from the associated research paper on cross-architecture distillation.
• Benchmarking knowledge distillation methods for passage ranking tasks.
• Investigating the transfer of ranking knowledge between different neural network architectures.

Strengths

• Data is directly linked to a peer-reviewed arXiv paper (2010.02666) with a public GitHub repository for documentation.
• Based on the established MSMARCO-Passage benchmark collection for information retrieval.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file size are unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Sebastian Hofstätter, Sophia Althammer, Michael Schröder, Mete Sertkan, Allan Hanbury (TU Wien)

Collection Method

Created for the associated research paper, likely derived from the MSMARCO-Passage collection.

Time Range

The paper was published on arXiv in October 2020.

Freshness

Last update date is unknown; freshness unverified.

Geography

null