domains / families / research / uai-reproducibility

uai-reproducibility

separate

Use when hardening reproducibility evidence for a UAI paper, including seeds, sampler convergence diagnostics, ELBO and calibration traces, dataset and hyperparameter disclosure, compute reporting, and honest code-availability statements, since UAI strongly encourages released code and data and reviews whether claims are convincingly backed.

When to use

• hardening reproducibility evidence for a UAI paper
• reporting seeds and sampler convergence diagnostics
• disclosing ELBO and calibration traces
• documenting dataset and hyperparameter disclosure
• reporting compute and code-availability statements
• auditing stochastic-inference results
• creating a determinism ledger
• performing a pre-submission reproducibility drill

When not to use

• tasks not related to UAI paper reproducibility

Skill page

Source