Final project presentation rubric (100 points)

Each team presents for 12 minutes and answers questions for 4 minutes. The talk is 10% of the course grade. Slides from every team are due through the Canvas assignment by 1:00 PM on Tuesday, April 13, whichever day the team presents.

Each criterion is worth 20 points and is scored in one of four bands. All members of a team receive the team grade.

Criterion	Excellent (18-20)	Good (14-17)	Fair (10-13)	Weak (0-9)
Problem and motivation The open problem, why it matters, what is new	The audience understands the open problem, its clinical or technical stakes, and what this study adds, within the first two minutes.	Problem and contribution are clear; motivation is thin or generic.	The problem is stated, but the contribution is unclear or confused with prior work.	The audience cannot tell what problem the team worked on.
Methods and baseline Data, model, and an honest comparison	Data (source, scale, splits), model, and design choices are explained correctly, with a baseline that makes the comparison fair.	Correct and complete; the baseline is reasonable but lightly justified.	Superficial or partly incorrect, or the baseline is too weak to support the claim.	No baseline, or methods that do not match the stated problem.
Evaluation and results Beyond one aggregate metric	Results go beyond a single aggregate number (for example calibration, subgroups, or a shift or robustness test) and are interpreted, including what they do not show.	Main results with some interpretation; one dimension beyond accuracy.	Results shown with little interpretation, or a single metric only.	Results missing, misreported, or not tied to the question.
Failure analysis and limitations What is shown versus what is claimed	Concrete failure analysis with examples, and limitations the team found themselves, stated plainly.	Some failure analysis; limitations mostly generic.	Limitations listed without analysis.	Neither, or claims that the results do not support.
Delivery Structure, slides, timing, questions	Clear structure and figures, finishes on time, every member speaks, questions answered directly.	Clear overall; minor timing or slide issues.	Hard to follow, text-heavy slides, timing clearly off, or uneven speaking time.	Disorganized, slides read aloud, or major timing problems.

The talk is judged on its own, not on the written paper, which has its own rubric in README.md. A study whose result is negative can score full marks: what is graded is the quality of the question, the comparison, and the honesty of the reporting.