domains / families / research / emnlp-reproducibility

emnlp-reproducibility

separate

Use when hardening an EMNLP paper's reproducibility record — answering the Responsible NLP checklist truthfully under desk-reject enforcement, pinning model versions and API snapshot dates, logging decoding parameters and prompts, documenting data licensing and annotation, and reporting compute so another lab could rerun the study.

When to use

• answering the Responsible NLP checklist
• pinning model versions and API snapshot dates
• logging decoding parameters and prompts
• documenting data licensing and annotation
• reporting compute and cost
• defining release honesty levels
• mapping paper numbers to run IDs

When not to use

• writing ritual limitations text that admits nothing
• claiming reproducibility without supporting evidence
• releasing datasets as bare zips without documentation

Skill page

Source