domains / families / research / neurips-experiments

neurips-experiments

separate

Use when stress-testing NeurIPS experimental evidence, including baselines, ablations, data splits, compute, negative results, real-world use, and claim-to-evidence calibration.

When to use

• stress-testing NeurIPS experimental evidence
• auditing baselines and ablations
• checking data splits and compute disclosure
• calibrating claims to evidence
• review-dimension stress testing
• rebuttal triage and preparation

When not to use

• winning a leaderboard without scientific meaning
• application reports with no ML contribution
• silently omitting the strongest comparison

Skill page

Source