Assessment
Grading, the paper presentation criteria, the paper replication project and every dated event of the 2025 offering in one place.
Dates
| Date | Event |
|---|---|
| 03 Jan 25 | Course begins: introduction and logistics |
| 10 Jan 25 | Class cancelled |
| 17 Jan 25 | Paper presentations |
| 22 Jan 25 | Paper presentations |
| 24 Jan 25 | Paper presentations |
| 29 Jan 25 | Guest talk: Dr. Tavpritesh Sethi |
| 31 Jan 25 | Paper presentations |
| 05 Feb 25 | Guest talk: Dr. Anupam Sobti |
| 19 Feb 25 | Project evaluation 1: proposal and Gantt chart |
| 20 Feb 25 | Guest talk: Alok, Google DeepMind |
The second project evaluation came after the mid-semester break and the final presentation at the end of the semester.
Grading
Everything is out of 100 marks.
| Component | Marks | How |
|---|---|---|
| Paper presentation | 10 | A 20-minute presentation of a paper, graded by the instructor and by peers. See criteria. |
| Paper review 1 | 5 | In class: based on the abstract, the introduction and fatal flaws. |
| Paper review 2 | 15 | At home, on a review platform: everything in review 1 plus a full AI paper review covering risks, ethics and datasets. Read existing OpenReview reviews first. |
| Class participation | 5 | 1 mark for each active class participation. |
| Guest lecture quiz | 5 | An in-class quiz on the guest lectures. |
| Project proposal and literature survey | 10 | Project evaluation 1. See the project. |
| Project mid-term presentation | 10 | Project evaluation 2. |
| Final project presentation | 15 | Project evaluation 3. |
| Attendance | 5 | |
| Assignment 1 | 10 | |
| Assignment 2 | 10 | |
| Total | 100 |
Paper presentation
Each individual paper presentation is marked out of 10.
| Criterion | Marks |
|---|---|
| Presentation quality (spoken and slides) | 3 |
| Q&A | 1 |
| Discussion: reasons to accept | 2 |
| Discussion: reasons to reject | 2 |
| Explaining the approach | 2 |
| Total | 10 |
Each presenter is expected to explain the paper’s key ideas clearly, defend their interpretation during Q&A, and contribute to the discussion on whether the paper should be accepted or rejected.
Paper replication project
Replicate a NeurIPS-like paper in a social good setting. The project is worth 35 marks over three group evaluations.
| Evaluation | Marks | Format and criteria |
|---|---|---|
| 1. Proposal and Gantt chart 19 Feb 2025 (Wednesday) | 10 | A 10-minute group presentation. Understanding of the paper (4): the paper and its key contributions. Feasibility analysis (3): whether the paper can be replicated with the available code and data. Gantt chart and planning (3): a clear timeline with milestones. |
| 2. Mid-project progress After the mid-semester break | 10 | A short update presentation of about 10 minutes. Replication progress (5): how much of the paper has been replicated. Challenges and solutions (3): problems faced and how the team is addressing them. Adherence to plan (2): whether the project is on track with the Gantt chart. |
| 3. Final presentation End of semester | 15 | A detailed group presentation with results and learnings. Successful replication (6): whether the key experiments and results were reproduced. Social good adaptation (5): how well the replication is set in a social impact context. Analysis and insights (4): interpretation of results, limitations and future scope. |