Project Rubric

Criterion	Weight within project	Excellent evidence
Research question & motivation	10%	precise, meaningful, scoped
Related work & baseline	10%	fair comparison, correct positioning
Technical method	20%	correct, inspectable, well-justified
Experimental design	20%	appropriate subjects, metrics, controls
Results & error analysis	15%	complete, transparent, interpretable
Reproducibility artifact	15%	rerunnable, versioned, documented
Report & presentation	10%	concise, technically persuasive

Penalties / failure modes

Severe problems include:

• no meaningful baseline;
• unrecoverable subject-system versions;
• evaluating only cherry-picked successful cases;
• invalid transformations included as quality improvements;
• train/test leakage;
• missing disclosure of material AI/model assistance;
• results that cannot be regenerated from the submitted artifact.