A Template for Research Talks

Every component of the slide theme, on one deck

Amir Masoud Nourollah

Cardiff University

Overview

  1. Text and emphasis
  2. Columns and cards
  3. Figures and tables
  4. Headline numbers and takeaways

Text and Emphasis

Write titles as claims, not topics

A lead sentence states the point of the slide in plain words.

  • Body text is large enough to read from the back of the room
  • Use bold for key terms, amber for your result, and teal for the baseline
  • Nested points stay readable
    • but keep them short

Key point: one sentence the audience should remember from this slide.

Warning: use amber for the surprising or failing case.

Columns and Cards

Two columns

Seen in training

  • Familiar chain lengths
  • Familiar functional groups

New combinations

  • Same parts, new arrangement
  • Errors grow by orders of magnitude

Three columns

Fusion

Two groups merge into one composite group.

Duplication

A familiar group appears twice.

Combination

Two known groups share one backbone.

Figures and Tables

A figure with a short explanation

The GMD benchmark at a glance.

Put the evidence on the left, and two or three lines on the right telling the audience what to look at.

Point at one thing only.

References go in a .slide-ref block and sit just above the footer.

A clean results table

Model Seen (meV/Å) Unseen (meV/Å) Ratio
SchNet 12.4 418.0 34×
MACE 3.1 96.2 31×
EquiFormerV2 2.7 88.5 33×
UMA (FT) 3.4 140.9 41×

Illustrative numbers only.

Numbers and Takeaways

1000× larger error on molecules built from familiar parts

Maths and code

The energy is a sum of atomic contributions:

\[ E = \sum_{i=1}^{N} \varepsilon\left(\mathbf{r}_i, \{\mathbf{r}_j\}_{j \in \mathcal{N}(i)}\right) \]

model = MACE.from_pretrained("mace-mp-0")
forces = model(atoms).forces

Thank you

Questions are welcome.

nourollah.me · arXiv:2605.08988