SliToPro.
← Gallery
Academic Paper

Grand Rounds Dossier — Clinical Case Two-Pane

A journal-formatted grand-rounds slide dividing patient history from diagnostic imaging in a strict two-column layout, with a numbered differential list and a JAMA-style case header. Built for radiology and internal-medicine rounds where clinical evidence must live beside narrative without competing.

This template refuses decoration: every element serves legibility or clinical hierarchy, nothing else. It exists for rooms where a missed finding has consequences — where body-serif type signals the authority of a peer-reviewed case report, a grayscale imaging panel commands undivided attention, and a single muted teal accent directs the eye to the differential without upstaging the image. It rejects brand gradients, rounded UI cards, and anything that telegraphs 'tech startup'; it should feel like a JAMA spread projected on a hospital wall at 6 a.m. rounds.

Palette — Nordic Frost

Prompt — Preview 1

Open in:ClaudeChatGPTGemini

Good for

  • · Grand-rounds and case-conference presentations with paired imaging
  • · Radiology or pathology teaching slides requiring image-label clarity
  • · Journal club case discussions in JAMA or NEJM format
  • · Fellowship didactics linking patient history to a single diagnostic image

Avoid for

  • · Statistical-results slides requiring multiple layered data charts
  • · Public-health awareness or patient-facing education decks
  • · Non-clinical academic disciplines with no imaging panel content
  • · Keynote or plenary talks without a discrete patient-case structure

Tags

#grand-rounds#two-pane#radiology#serif-body#clinical-case#differential-list

More in Academic Paper

Vol. 137 · No. 4 · Spring Term · Casebook Ivory Law Review Hartwell v. Commonwealth, 412 U.S. 318, 329 (1974) cited in Restatement (Third) of Torts § 14 cmt. b (Am. Law Inst. 2010) HOLDING · PART III · SUBSECTION (B) ¶ 27. A duty of reasonable care arises where a defendant assumes affirmative control over a foreseeable risk. Argued by Jane Doe, Counsel of Record · Brief prepared with OOO & Associates SUBMITTED · 24 APRIL 2026 · OCTOBER TERM SITTING — 412 — CASEBOOK IVORY · §03

Casebook Ivory — Bluebook Holding Statement

A law-journal slide structured around a Bluebook citation header and bold-serif holding statement, with numbered argument paragraphs, superscript footnote markers, and a hairline-ruled footnote block that mirrors the typographic discipline of a printed Harvard Law Review page.

Academic Paper
Lecture Notes · Series III · Analysis Seminar On the Convergence of Sobolev Embeddings in Fractional Domains A theorem–definition scaffold with corollaries on compact operators [Your Name] Department of Mathematics, Institute for Analytic Studies 24 April 2026 Vol. XIV § No. 02 MSC 46E35 · 35J60 pre-print · circulating draft

Computer Modern — Theorem & Definition Scaffold

A LaTeX-faithful lecture canvas for math-heavy content: a thin royal-blue-ruled definition box at top, italic serif theorem statement, and parenthesized numbered proof steps — every glyph set in Computer Modern, exactly as the TeX tradition demands.

Academic Paper
PREPRINT · UNDER REVIEW AS A CONFERENCE PAPER AT NEURIPS 2026 Spectral Consistency Regularization for Low-Rank Adaptation in Diffusion Transformers Jane Doe1 A. Researcher1,2 B. Collaborator2 C. Reviewer3 D. Advisor1 1 Department of Computer Science, Example University 2 Institute for Machine Intelligence 3 Studio Research Lab { jane.doe, a.researcher } @ example.edu ABSTRACT We introduce Spectral Consistency Regularization (SCR), a training-time objective that constrains the singular-value spectrum of low-rank adapter updates to match a reference distribution induced by the pre-trained backbone. Given a frozen diffusion transformer fθ and a rank-r adapter ΔW = BA, we penalize the Wasserstein-2 distance between σ(ΔW) and the empirical spectrum of the corresponding backbone block, yielding a loss of the form ℒSCR = 𝔼[‖σ(BA) − σ*‖²]. The resulting adapters remain faithful to the base model's inductive priors while admitting task-specific deformation. Across four conditional generation benchmarks and three backbone scales (0.6B — 8.2B parameters), SCR recovers 94.1% of full fine-tuning FID at 0.41% of the trainable parameter budget, exceeding LoRA, DoRA, and orthogonal-init baselines by a consistent margin (Δ-FID ≥ 1.8). Ablations isolate the spectral term as the principal contributor and reveal that effective rank collapses monotonically without it. We release code, trained adapters, and evaluation protocols. Full derivations appear in Appendix A. Keywords — parameter-efficient fine-tuning · diffusion transformers · spectral regularization · low-rank adaptation · generative modeling arXiv:2604.01127v1 [cs.LG] 24 Apr 2026 NeurIPS 2026 · Submission #4412 · Page 1 / 18

Manuscript Grid — NeurIPS Two-Column Folio

A LaTeX-faithful two-column slide modelled on NeurIPS and ICML paper conventions, with justified serif body, centred numbered equations, and an inset results figure — the only ML-result template that treats typographic protocol as a design feature, not a constraint.

Academic Paper