Paper presentations & reviews

There will be two presentation sessions for MSc students, and you will take a different role in each. The schedule gives this year’s dates.

Your presentation involves discussing the paper as though you were one of its authors. Your turn as Reviewer 2, the designated discussant, takes place during the week you’re not presenting, when you are assigned to review the paper from another talk. That role is live and spoken; the written referee report, later in the course and on another paper, is separate work.

Session format

Each slot runs about twenty minutes: ten for the talk, five for the assigned reviewer’s questions, and five for the floor and the lecturer.

Ten minutes is short. Choose one clear thread through the paper rather than trying to cover everything. You are marked on explaining the method well, not on completeness.

If you are presenting

Present as if you were one of the authors, presenting the work at a conference. Motivate the problem as they would, explain the method as their own work, and defend the design choices when questioned. You are not presenting your reproduction plan or your opinion of the paper: for twenty minutes, you are the person who wrote it and must defend it.

Criterion	Weight	What earns full marks
Motivation and framing	10%	Sets up the problem as the authors would: what was broken before, and why this was worth doing.
Method exposition	25%	The mechanism explained correctly and at the right level for a conference audience, with design choices motivated as the authors’ own.
Results and evidence	15%	Presents the experiments that actually support the claims, ablations included, without overselling.
Q&A and defence in role	30%	Answers questions that are not on a slide, defends design choices, and concedes limitations as a credible author would.
Slides and delivery	20%	Keeps to time, audible, slides legible. Slides do real work: the paper’s own figures rather than bullets describing them, the key method diagram redrawn or annotated so that it carries the explanation, and no walls of text or stock-icon filler.

You will take at least three questions, including at least one that goes beyond your slides.

If you are reviewing

Every presentation has a Reviewer 2: another student, assigned in advance, who has read the paper thoroughly and whose role in the session is to make the case against it. Not hostility, but the assigned scepticism a good reviewer brings. Prepare three questions the author will not enjoy, and be ready to say which of them you think the paper survives.

Go after what actually threatens the claims, such as a missing ablation, a favourable baseline, or an understated limitation, rather than definitions or trivia. Your five minutes are guaranteed and come before the floor, so the follow-up is yours to lead.

Do not share your questions with the presenter beforehand.

Criterion	Weight	What earns full marks
Evidence of having read the paper	25%	Questions come from the paper itself rather than only from the talk, and pick up something the presenter skipped or smoothed over.
Targets real weak spots	35%	Goes after what actually threatens the claims, not trivia or definitions.
Constructive framing	20%	Puts the question so that the presenter can engage with it rather than defend themselves. Probing, not point-scoring.
Engagement and follow-up	20%	Listens to the answer and pushes where it does not land, rather than reading out the next prepared question.

The two marks are independent. A Reviewer 2 who asks weak questions cannot cost the presenter, and a presenter who fumbles cannot cost their Reviewer 2, because presenters are marked on how they handle questions from the whole room, including the lecturer.

On using an LLM

Using an LLM to help prepare your slides is fine and expected. The final quality of the slides and presentation is your responsibility. Should it include too much text, or push you into a rigid, pre-set structure, that still costs you marks. Note where the marks sit: the largest single component is the live Q&A, which no amount of preparation can generate for you. Spend your effort on understanding the paper well enough to defend it.

Practicalities

Allocation. Who presents which paper, and who reviews it, is posted on Moodle before the first session. You present once and review once, in opposite weeks, and the pairing is offset so that no two students review each other. Paper titles there link to the version the presenter is working from, so you and your Reviewer 2 read the same text.

Before the session. Nothing is handed in. Bring your slides on the day. Reviewer 2’s questions are not submitted and are not sent to the presenter.

During the session. The session will be recorded for later reference if necessary.

After the session. Submit your slides to Moodle, as you presented them rather than tidied up afterwards. Marks and brief comments for both roles come back the same way.

Preparing. The two rubrics above are the brief: they say what earns full marks in each role. Reading a paper sets out the three passes; Reviewer 2 needs the second at least, and the third is what turns a polite question into one the author has to think about.

# Paper presentations & reviews {.unnumbered} ## Paper presentations & reviews There will be two presentation sessions for MSc students, and you will take a different role in each. The [schedule](../schedule.qmd) gives this year's dates. Your **presentation** involves discussing the paper as though you were one of its authors. Your turn as **Reviewer 2**, the designated discussant, takes place during the week you're not presenting, when you are assigned to review the paper from another talk. That role is live and spoken; the written referee report, later in the course and on another paper, is separate work. ## Session format Each slot runs about twenty minutes: ten for the talk, five for the assigned reviewer's questions, and five for the floor and the lecturer. Ten minutes is short. Choose one clear thread through the paper rather than trying to cover everything. You are marked on explaining the method well, not on completeness. ## If you are presenting Present as if you were one of the authors, presenting the work at a conference. Motivate the problem as they would, explain the method as their own work, and defend the design choices when questioned. You are not presenting your reproduction plan or your opinion of the paper: for twenty minutes, you are the person who wrote it and must defend it. | Criterion | Weight | What earns full marks | | ----------------------- | ------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Motivation and framing | 10% | Sets up the problem as the authors would: what was broken before, and why this was worth doing. | | Method exposition | 25% | The mechanism explained correctly and at the right level for a conference audience, with design choices motivated as the authors' own. | | Results and evidence | 15% | Presents the experiments that actually support the claims, ablations included, without overselling. | | Q&A and defence in role | 30% | Answers questions that are not on a slide, defends design choices, and concedes limitations as a credible author would. | | Slides and delivery | 20% | Keeps to time, audible, slides legible. Slides do real work: the paper's own figures rather than bullets describing them, the key method diagram redrawn or annotated so that it carries the explanation, and no walls of text or stock-icon filler. | : {tbl-colwidths="[24,10,66]"} You will take at least three questions, including at least one that goes beyond your slides. ## If you are reviewing Every presentation has a Reviewer 2: another student, assigned in advance, who has read the paper thoroughly and whose role in the session is to make the case against it. Not hostility, but the assigned scepticism a good reviewer brings. Prepare three questions the author will not enjoy, and be ready to say which of them you think the paper survives. Go after what actually threatens the claims, such as a missing ablation, a favourable baseline, or an understated limitation, rather than definitions or trivia. Your five minutes are guaranteed and come before the floor, so the follow-up is yours to lead. **Do not share your questions with the presenter beforehand.** | Criterion | Weight | What earns full marks | | --------------------------------- | ------ | ---------------------------------------------------------------------------------------------------------------------------------- | | Evidence of having read the paper | 25% | Questions come from the paper itself rather than only from the talk, and pick up something the presenter skipped or smoothed over. | | Targets real weak spots | 35% | Goes after what actually threatens the claims, not trivia or definitions. | | Constructive framing | 20% | Puts the question so that the presenter can engage with it rather than defend themselves. Probing, not point-scoring. | | Engagement and follow-up | 20% | Listens to the answer and pushes where it does not land, rather than reading out the next prepared question. | : {tbl-colwidths="[24,10,66]"} The two marks are independent. A Reviewer 2 who asks weak questions cannot cost the presenter, and a presenter who fumbles cannot cost their Reviewer 2, because presenters are marked on how they handle questions from the whole room, including the lecturer. ## On using an LLM Using an LLM to help prepare your slides is fine and expected. The final quality of the slides and presentation is your responsibility. Should it include too much text, or push you into a rigid, pre-set structure, that still costs you marks. Note where the marks sit: the largest single component is the live Q&A, which no amount of preparation can generate for you. Spend your effort on understanding the paper well enough to defend it. ## Practicalities **Allocation.** Who presents which paper, and who reviews it, is posted on Moodle before the first session. You present once and review once, in opposite weeks, and the pairing is offset so that no two students review each other. Paper titles there link to the version the presenter is working from, so you and your Reviewer 2 read the same text. **Before the session.** Nothing is handed in. Bring your slides on the day. Reviewer 2's questions are not submitted and are not sent to the presenter. **During the session.** The session will be recorded for later reference if necessary. **After the session.** Submit your slides to Moodle, as you presented them rather than tidied up afterwards. Marks and brief comments for both roles come back the same way. **Preparing.** The two rubrics above are the brief: they say what earns full marks in each role. [Reading a paper](reading.qmd) sets out the three passes; Reviewer 2 needs the second at least, and the third is what turns a polite question into one the author has to think about.