Prepare and defend a research presentation about one recent project, then connect its lessons to AI-agent systems. Use your own work; do not invent results, baselines, or ownership. ### Clarifying Questions to Ask - How much presentation time is available, and how much is reserved for questions? - Is the audience expected to know the research area already? - Should the discussion emphasize scientific novelty, engineering execution, or product relevance? ### Part 1 - Frame the Research Problem Explain the problem, why it matters, the prior limitation you addressed, and the hypothesis or research question. Define the evaluation target before presenting the method. #### What This Part Should Cover ```premium-lock What This Part Should Cover ``` ### Part 2 - Defend the Method and Evidence Describe the method at the level needed to evaluate it. Explain data selection, baselines, metrics, ablations, error analysis, and the most important result. Identify threats to validity and one result that changed your thinking. #### What This Part Should Cover ```premium-lock What This Part Should Cover ``` ### Part 3 - Connect the Work to AI Agents Discuss which parts of the project transfer to an agent setting, such as planning, tool use, memory, retrieval, evaluation, or safety. Also state where the analogy breaks and propose one experiment that would test the connection. #### What This Part Should Cover ```premium-lock What This Part Should Cover ``` ### What a Strong Answer Covers ```premium-lock What a Strong Answer Covers ``` ### Follow-up Questions 1. Which ablation would most likely overturn your conclusion? 2. How would you evaluate an agent when offline labels do not capture task success? 3. What would you change if compute or data were reduced by an order of magnitude?