Master's Thesis Proposal

MAI Program — Agentic AI & VLA · Supervisor: Prof. Anis Koubaa

1. Problem

In two or three sentences, state the concrete problem. What breaks today, and why does it matter?

2. Research Gap

What is missing in the current literature? Cite the two or three closest works and state precisely what they do not address.

3. Proposed Contribution

State exactly what you will build and what is new about it. Keep it specific and bounded for one student in twelve months.

4. Research Questions

1. RQ1 — the main measurable question.
2. RQ2 — a secondary question (ablation, generalization, cost, safety).

5. Methodology

Base model(s) and fine-tuning approach (LoRA/QLoRA), the agentic wrapper, and the pipeline. A simple diagram helps.

6. Datasets, Simulators & Tools

List the datasets/benchmarks and simulators, and confirm they are public and fit the GPU budget.

7. Evaluation Metrics & Baselines

Define the metrics (e.g., success rate, precision/recall, pass^k, cost, energy) and the baselines you will compare against, including at least one strong frontier-model baseline.

8. Timeline (12 months)

• Months 1–2: literature review, environment setup, baseline reproduction.
• Months 3–6: build the method; first experiments.
• Months 7–9: full evaluation, ablations, robustness.
• Months 10–12: writing, submission, thesis defense.

9. Expected Contribution & Publication Plan

State the expected scientific contribution and the concrete venue and submission window you are targeting.

10. References

Five to ten key references, including the closest prior works named in the Gap section.