AI Graphics for Posters

Full rules, worked examples, and the mandatory review gates for generating poster visuals. SKILL.md carries the hard limits; this file carries the reasoning, the per-graphic-type tables, and the complete before/after prompt examples.

CRITICAL: Preventing Content Overflow

⚠️ POSTERS MUST NOT HAVE TEXT OR CONTENT CUT OFF AT EDGES.

Common Overflow Problems:

Title/footer text extending beyond page boundaries
Too many sections crammed into available space
Figures placed too close to edges
Text blocks exceeding column widths

Prevention Rules:

1. Limit Content Sections (MAXIMUM 5-6 sections for A0):

✅ GOOD - 5 sections with room to breathe:
   - Title/Header
   - Introduction/Problem
   - Methods
   - Results (1-2 key findings)
   - Conclusions

❌ BAD - 8+ sections crammed together:
   - Overview, Introduction, Background, Methods, 
   - Results 1, Results 2, Discussion, Conclusions, Future Work

2. Set Safe Margins in LaTeX:

latex

% tikzposter - add generous margins
\documentclass[25pt, a0paper, portrait, margin=25mm]{tikzposter}

% baposter - ensure content doesn't touch edges
\begin{poster}{
  columns=3,
  colspacing=2em,           % Space between columns
  headerheight=0.1\textheight,  % Smaller header
  % Leave space at bottom
}

3. Figure Sizing - Never 100% Width:

latex

% Leave margins around figures
\includegraphics[width=0.85\linewidth]{figure.png}  % NOT 1.0\linewidth

4. Check for Overflow Before Printing:

bash

# Compile and check PDF at 100% zoom
pdflatex poster.tex

# Look for:
# - Text cut off at any edge
# - Content touching page boundaries  
# - Overfull hbox warnings in .log file
grep -i "overfull" poster.log

5. Word Count Limits:

A0 poster: 300-800 words MAXIMUM
Per section: 50-100 words maximum
If you have more content: Cut it or make a handout

CRITICAL: Poster-Size Font Requirements

⚠️ ALL text within AI-generated visualizations MUST be poster-readable.

When generating graphics for posters, you MUST include font size specifications in EVERY prompt. Poster graphics are viewed from 4-6 feet away, so text must be LARGE.

⚠️ COMMON PROBLEM: Content Overflow and Density

The #1 issue with AI-generated poster graphics is TOO MUCH CONTENT. This causes:

Text overflow beyond boundaries
Unreadable small fonts
Cluttered, overwhelming visuals
Poor white space usage

SOLUTION: Generate SIMPLE graphics with MINIMAL content.

MANDATORY prompt requirements for EVERY poster graphic:

POSTER FORMAT REQUIREMENTS (STRICTLY ENFORCE):
- ABSOLUTE MAXIMUM 3-4 elements per graphic (3 is ideal)
- ABSOLUTE MAXIMUM 10 words total in the entire graphic
- NO complex workflows with 5+ steps (split into 2-3 simple graphics instead)
- NO multi-level nested diagrams (flatten to single level)
- NO case studies with multiple sub-sections (one key point per case)
- ALL text GIANT BOLD (80pt+ for labels, 120pt+ for key numbers)
- High contrast ONLY (dark on white OR white on dark, NO gradients with text)
- MANDATORY 50% white space minimum (half the graphic should be empty)
- Thick lines only (5px+ minimum), large icons (200px+ minimum)
- ONE SINGLE MESSAGE per graphic (not 3 related messages)

⚠️ BEFORE GENERATING: Review your prompt and count elements

If your description has 5+ items → STOP. Split into multiple graphics
If your workflow has 5+ stages → STOP. Show only 3-4 high-level steps
If your comparison has 4+ methods → STOP. Show only top 3 or Our vs Best Baseline

Content limits per graphic type (STRICT):

Graphic Type	Max Elements	Max Words	Reject If	Good Example
Flowchart	3-4 boxes MAX	8 words	5+ stages, nested steps	"DISCOVER → VALIDATE → APPROVE" (3 words)
Key findings	3 items MAX	9 words	4+ metrics, paragraphs	"95% ACCURATE" "2X FASTER" "FDA READY" (6 words)
Comparison chart	3 bars MAX	6 words	4+ methods, legend text	"OURS: 95%" "BEST: 85%" (4 words)
Case study	1 case, 3 elements	6 words	Multiple cases, substories	Logo + "18 MONTHS" + "to discovery" (2 words)
Timeline	3-4 points MAX	8 words	Year-by-year detail	"2020 START" "2022 TRIAL" "2024 APPROVED" (6 words)

Example - WRONG (7-stage workflow - TOO COMPLEX):

bash

# ❌ BAD - This creates tiny unreadable text like the drug discovery poster
python scripts/generate_schematic.py "Drug discovery workflow showing: Stage 1 Target Identification, Stage 2 Molecular Synthesis, Stage 3 Virtual Screening, Stage 4 AI Lead Optimization, Stage 5 Clinical Trial Design, Stage 6 FDA Approval. Include success metrics, timelines, and validation steps for each stage." -o figures/workflow.png
# Result: 7+ stages with tiny text, unreadable from 6 feet - POSTER FAILURE

Example - CORRECT (simplified to 3 key stages):

bash

# ✅ GOOD - Same content, split into ONE simple high-level graphic
python scripts/generate_schematic.py "POSTER FORMAT for A0. ULTRA-SIMPLE 3-box workflow: 'DISCOVER' → 'VALIDATE' → 'APPROVE'. Each word in GIANT bold (120pt+). Thick arrows (10px). 60% white space. NO substeps, NO details. 3 words total. Readable from 10 feet." -o figures/workflow_overview.png
# Result: Clean, impactful, readable - can add detail graphics separately if needed

Example - WRONG (complex case studies with multiple sections):

bash

# ❌ BAD - Creates cramped unreadable sections
python scripts/generate_schematic.py "Case studies: Insilico Medicine (drug candidate, discovery time, clinical trials), Recursion Pharma (platform, methodology, results), Exscientia (drug candidates, FDA status, timeline). Include company logos, metrics, and outcomes." -o figures/cases.png
# Result: 3 case studies with 4+ elements each = 12+ total elements, tiny text

Example - CORRECT (one case study, one key metric):

bash

# ✅ GOOD - Show ONE case with ONE key number
python scripts/generate_schematic.py "POSTER FORMAT for A0. ONE case study card: Company logo (large), '18 MONTHS' in GIANT text (150pt), 'to discovery' below (60pt). 3 elements total: logo + number + caption. 50% white space. Readable from 10 feet." -o figures/case_single.png
# Result: Clear, readable, impactful. Make 3 separate graphics if you need 3 cases.

Example - WRONG (key findings too complex):

bash

# BAD - too many items, too much detail
python scripts/generate_schematic.py "Key findings showing 8 metrics: accuracy 95%, precision 92%, recall 94%, F1 0.93, AUC 0.97, training time 2.3 hours, inference 50ms, model size 145MB with comparison to 5 baseline methods" -o figures/findings.png
# Result: Cramped graphic with tiny numbers

Example - CORRECT (key findings simple):

bash

# GOOD - only 3 key items, giant numbers
python scripts/generate_schematic.py "POSTER FORMAT for A0. KEY FINDINGS with ONLY 3 large cards. Card 1: '95%' in GIANT text (120pt) with 'ACCURACY' below (48pt). Card 2: '2X' in GIANT text with 'FASTER' below. Card 3: checkmark icon with 'VALIDATED' in large text. 50% white space. High contrast colors. NO other text or details." -o figures/findings.png
# Result: Bold, readable impact statement

Font size reference for poster prompts:

Element	Minimum Size	Prompt Keywords
Main numbers/metrics	72pt+	"huge", "very large", "giant", "poster-size"
Section titles	60pt+	"large bold", "prominent"
Labels/captions	36pt+	"readable from 6 feet", "clear labels"
Body text	24pt+	"poster-readable", "large text"

Always include in prompts:

"POSTER FORMAT" or "for A0 poster" or "readable from 6 feet"
"VERY LARGE TEXT" or "huge bold fonts"
Specific text that should appear (so it's baked into the image)
"minimal text, maximum impact"
"high contrast" for readability
"generous margins" and "no text near edges"

CRITICAL: AI-Generated Graphic Sizing

⚠️ Each AI-generated graphic should focus on ONE concept with MINIMAL content.

Problem: Generating complex diagrams with many elements leads to small text.

Solution: Generate SIMPLE graphics with FEW elements and LARGE text.

Example - WRONG (too complex, text will be small):

bash

# BAD - too many elements in one graphic
python scripts/generate_schematic.py "Complete ML pipeline showing data collection, 
preprocessing with 5 steps, feature engineering with 8 techniques, model training 
with hyperparameter tuning, validation with cross-validation, and deployment with 
monitoring. Include all labels and descriptions." -o figures/pipeline.png

Example - CORRECT (simple, focused, large text):

bash

# GOOD - split into multiple simple graphics with large text

# Graphic 1: High-level overview (3-4 elements max)
python scripts/generate_schematic.py "POSTER FORMAT for A0: Simple 4-step pipeline. 
Four large boxes: DATA → PROCESS → MODEL → RESULTS. 
GIANT labels (80pt+), thick arrows, lots of white space. 
Only 4 words total. Readable from 8 feet." -o figures/overview.png

# Graphic 2: Key result (1 metric highlighted)
python scripts/generate_schematic.py "POSTER FORMAT for A0: Single key metric display.
Giant '95%' text (150pt+) with 'ACCURACY' below (60pt+).
Checkmark icon. Minimal design, high contrast.
Readable from 10 feet." -o figures/accuracy.png

Rules for AI-generated poster graphics:

Rule	Limit	Reason
Elements per graphic	3-5 maximum	More elements = smaller text
Words per graphic	10-15 maximum	Minimal text = larger fonts
Flowchart steps	4-5 maximum	Keeps labels readable
Chart categories	3-4 maximum	Prevents crowding
Nested levels	1-2 maximum	Avoids complexity

Split complex content into multiple simple graphics:

Instead of 1 complex diagram with 12 elements:
→ Create 3 simple diagrams with 4 elements each
→ Each graphic can have LARGER text
→ Arrange in poster with clear visual flow

Step 0: MANDATORY Pre-Generation Review (DO THIS FIRST)

⚠️ BEFORE generating ANY graphics, review your content plan:

For EACH planned graphic, ask these questions:

Element count: Can I describe this in 3-4 items or less?
- ❌ NO → Simplify or split into multiple graphics
- ✅ YES → Continue
Complexity check: Is this a multi-stage workflow (5+ steps) or nested diagram?
- ❌ YES → Flatten to 3-4 high-level steps only
- ✅ NO → Continue
Word count: Can I describe all text in 10 words or less?
- ❌ NO → Cut text, use single-word labels
- ✅ YES → Continue
Message clarity: Does this graphic convey ONE clear message?
- ❌ NO → Split into multiple focused graphics
- ✅ YES → Continue to generation

Common patterns that ALWAYS fail (reject these):

"Show stages 1 through 7..." → Split into high-level overview (3 stages) + detail graphics
"Multiple case studies..." → One case per graphic
"Timeline from 2015 to 2024 with annual milestones..." → Show only 3-4 key years
"Comparison of 6 methods..." → Show only top 3 or Our method vs Best baseline
"Architecture with all layers and connections..." → High-level only (3-4 components)

Step 1: Plan Your Poster Elements

After passing the pre-generation review, identify visual elements needed:

Title Block - Stylized title with institutional branding (optional - can be LaTeX text)
Introduction Graphic - Conceptual overview (3 elements max)
Methods Diagram - High-level workflow (3-4 steps max)
Results Figures - Key findings (3 metrics max per figure, may need 2-3 separate figures)
Conclusion Graphic - Summary visual (3 takeaways max)
Supplementary Icons - Simple icons, QR codes, logos (minimal)

Step 2: Generate Each Element (After Pre-Generation Review)

⚠️ CRITICAL: Review Step 0 checklist before proceeding.

Use the appropriate tool for each element type:

For Schematics and Diagrams (scientific-schematics):

bash

# Create figures directory
mkdir -p figures

# Drug discovery workflow - HIGH-LEVEL ONLY, 3 stages
# BAD: "Stage 1: Target ID, Stage 2: Molecular Synthesis, Stage 3: Virtual Screening, Stage 4: AI Lead Opt..."
# GOOD: Collapse to 3 mega-stages
python scripts/generate_schematic.py "POSTER FORMAT for A0. ULTRA-SIMPLE 3-box workflow: 'DISCOVER' (120pt bold) → 'VALIDATE' (120pt bold) → 'APPROVE' (120pt bold). Thick arrows (10px). 60% white space. ONLY these 3 words. NO substeps. Readable from 12 feet." -o figures/workflow_simple.png

# System architecture - MAXIMUM 3 components
python scripts/generate_schematic.py "POSTER FORMAT for A0. ULTRA-SIMPLE 3-component stack: 'DATA' box (120pt) → 'AI MODEL' box (120pt) → 'PREDICTION' box (120pt). Thick vertical arrows. 60% white space. 3 words only. Readable from 12 feet." -o figures/architecture.png

# Timeline - ONLY 3 key milestones (not year-by-year)
# BAD: "2018, 2019, 2020, 2021, 2022, 2023, 2024 with events"
# GOOD: Only 3 breakthrough moments
python scripts/generate_schematic.py "POSTER FORMAT for A0. Timeline with ONLY 3 points: '2018' + icon, '2021' + icon, '2024' + icon. GIANT years (120pt). Large icons. 60% white space. NO connecting lines or details. Readable from 12 feet." -o figures/timeline.png

# Case study - ONE case, ONE key metric
# BAD: "3 case studies: Insilico (details), Recursion (details), Exscientia (details)"
# GOOD: ONE case with ONE number
python scripts/generate_schematic.py "POSTER FORMAT for A0. ONE case study: Large logo + '18 MONTHS' (150pt bold) + 'to discovery' (60pt). 3 elements total. 60% white space. Readable from 12 feet." -o figures/case1.png

# If you need 3 cases → make 3 separate simple graphics (not one complex graphic)

For Stylized Blocks and Graphics (Nano Banana Pro):

bash

# Title block - SIMPLE
python scripts/generate_schematic.py "POSTER FORMAT for A0. Title block: 'ML FOR DRUG DISCOVERY' in HUGE bold text (120pt+). Dark blue background. ONE subtle icon. NO other text. 40% white space. Readable from 15 feet." -o figures/title_block.png

# Introduction visual - SIMPLE, 3 elements only
python scripts/generate_schematic.py "POSTER FORMAT for A0. SIMPLE problem visual with ONLY 3 icons: drug icon, arrow, target icon. ONE label per icon (80pt+). 50% white space. NO detailed text. Readable from 8 feet." -o figures/intro_visual.png

# Conclusion/summary - ONLY 3 items, GIANT numbers
python scripts/generate_schematic.py "POSTER FORMAT for A0. KEY FINDINGS with EXACTLY 3 cards only. Card 1: '95%' (150pt font) with 'ACCURACY' (60pt). Card 2: '2X' (150pt) with 'FASTER' (60pt). Card 3: checkmark icon with 'READY' (60pt). 50% white space. NO other text. Readable from 10 feet." -o figures/conclusions_graphic.png

# Background visual - SIMPLE, 3 icons only
python scripts/generate_schematic.py "POSTER FORMAT for A0. SIMPLE visual with ONLY 3 large icons in a row: problem icon → challenge icon → impact icon. ONE word label each (80pt+). 50% white space. NO detailed text. Readable from 8 feet." -o figures/background_visual.png

For Data Visualizations - SIMPLE, 3 bars max:

bash

# SIMPLE chart with ONLY 3 bars, GIANT labels
python scripts/generate_schematic.py "POSTER FORMAT for A0. SIMPLE bar chart with ONLY 3 bars: BASELINE (70%), EXISTING (85%), OURS (95%). GIANT percentage labels ON the bars (100pt+). NO axis labels, NO legend, NO gridlines. Our bar highlighted in different color. 40% white space. Readable from 8 feet." -o figures/comparison_chart.png

Step 2b: MANDATORY Post-Generation Review (Before Assembly)

⚠️ CRITICAL: Review EVERY generated graphic before adding to poster.

For each generated figure, open at 25% zoom and check:

✅ PASS criteria (all must be true):
- Can read ALL text clearly at 25% zoom
- Count elements: 3-4 or fewer
- White space: 50%+ of image is empty
- Simple enough to understand in 2 seconds
- NOT a complex workflow with 5+ stages
- NOT multiple nested sections
❌ FAIL criteria (regenerate if ANY are true):
- Text is small or hard to read at 25% zoom → REGENERATE with "150pt+" fonts
- More than 4 elements → REGENERATE with "ONLY 3 elements"
- Less than 50% white space → REGENERATE with "60% white space"
- Complex multi-stage workflow → SPLIT into 2-3 simple graphics
- Multiple case studies cramped together → SPLIT into separate graphics
- Takes more than 3 seconds to understand → SIMPLIFY and regenerate

Common failures and fixes:

"7-stage workflow with tiny text" → Regenerate as "3 high-level stages only"
"3 case studies in one graphic" → Generate 3 separate simple graphics
"Timeline with 8 years" → Regenerate with "ONLY 3 key milestones"
"Comparison of 5 methods" → Regenerate with "ONLY Our method vs Best baseline (2 bars)"

DO NOT PROCEED to assembly if ANY graphic fails the checks above.

Step 3: Assemble in LaTeX Template

After all figures pass the post-generation review, include them in your poster template:

tikzposter example:

latex

\documentclass[25pt, a0paper, portrait]{tikzposter}

\begin{document}

\maketitle

\begin{columns}
\column{0.5}

\block{Introduction}{
  \centering
  \includegraphics[width=0.85\linewidth]{figures/intro_visual.png}
  
  \vspace{0.5em}
  Brief context text here (2-3 sentences max).
}

\block{Methods}{
  \centering
  \includegraphics[width=0.9\linewidth]{figures/methods_flowchart.png}
}

\column{0.5}

\block{Results}{
  \begin{minipage}{0.48\linewidth}
    \centering
    \includegraphics[width=\linewidth]{figures/result_1.png}
  \end{minipage}
  \hfill
  \begin{minipage}{0.48\linewidth}
    \centering
    \includegraphics[width=\linewidth]{figures/result_2.png}
  \end{minipage}
  
  \vspace{0.5em}
  Key findings in 3-4 bullet points.
}

\block{Conclusions}{
  \centering
  \includegraphics[width=0.8\linewidth]{figures/conclusions_graphic.png}
}

\end{columns}

\end{document}

baposter example:

latex

\headerbox{Methods}{name=methods,column=0,row=0}{
  \centering
  \includegraphics[width=0.95\linewidth]{figures/methods_flowchart.png}
}

\headerbox{Results}{name=results,column=1,row=0}{
  \includegraphics[width=\linewidth]{figures/comparison_chart.png}
  \vspace{0.3em}
  
  Key finding: Our method achieves 92% accuracy.
}

Example: Complete Poster Generation Workflow

Full workflow with ALL quality checks:

bash

# STEP 0: Pre-Generation Review (MANDATORY)
# Content plan: Drug discovery poster
# - Workflow: 7 stages → ❌ TOO MANY → Reduce to 3 mega-stages ✅
# - 3 case studies → ❌ TOO MANY → One case per graphic (make 3 graphics) ✅
# - Timeline 2018-2024 → ❌ TOO DETAILED → Only 3 key years ✅

# STEP 1: Create figures directory
mkdir -p figures

# STEP 2: Generate ULTRA-SIMPLE graphics with strict limits

# Workflow - HIGH-LEVEL ONLY (collapsed from 7 stages to 3)
python scripts/generate_schematic.py "POSTER FORMAT for A0. ULTRA-SIMPLE 3-box workflow: 'DISCOVER' → 'VALIDATE' → 'APPROVE'. Each word 120pt+ bold. Thick arrows (10px). 60% white space. ONLY 3 words total. Readable from 12 feet." -o figures/workflow.png

# Case study 1 - ONE case, ONE metric (will make 3 separate graphics)
python scripts/generate_schematic.py "POSTER FORMAT for A0. ONE case: Company logo + '18 MONTHS' (150pt bold) + 'to drug discovery' (60pt). 3 elements only. 60% white space. Readable from 12 feet." -o figures/case1.png

python scripts/generate_schematic.py "POSTER FORMAT for A0. ONE case: Company logo + '95% SUCCESS' (150pt bold) + 'in trials' (60pt). 3 elements only. 60% white space." -o figures/case2.png

python scripts/generate_schematic.py "POSTER FORMAT for A0. ONE case: Company logo + 'FDA APPROVED' (150pt bold) + '2024' (60pt). 3 elements only. 60% white space." -o figures/case3.png

# Timeline - ONLY 3 key years (not 7 years)
python scripts/generate_schematic.py "POSTER FORMAT for A0. ONLY 3 years: '2018' (150pt) + icon, '2021' (150pt) + icon, '2024' (150pt) + icon. Large icons. 60% white space. NO lines or details. Readable from 12 feet." -o figures/timeline.png

# Results - ONLY 2 bars (our method vs best baseline, not 5 methods)
python scripts/generate_schematic.py "POSTER FORMAT for A0. TWO bars only: 'BASELINE 70%' and 'OURS 95%' (highlighted). GIANT percentages (150pt) ON bars. NO axis, NO legend. 60% white space. Readable from 12 feet." -o figures/results.png

# STEP 2b: Post-Generation Review (MANDATORY)
# Open each figure at 25% zoom:
# ✅ workflow.png: 3 elements, text readable, 60% white - PASS
# ✅ case1.png: 3 elements, giant numbers, clean - PASS
# ✅ case2.png: 3 elements, giant numbers, clean - PASS  
# ✅ case3.png: 3 elements, giant numbers, clean - PASS
# ✅ timeline.png: 3 elements, readable, simple - PASS
# ✅ results.png: 2 bars, giant percentages, clear - PASS
# ALL PASS → Proceed to assembly

# STEP 3: Compile LaTeX poster
pdflatex poster.tex

# STEP 4: PDF Overflow Check (see Section 11)
grep "Overfull" poster.log
# Open at 100% and check all 4 edges

If ANY graphic fails Step 2b review:

Too many elements → Regenerate with "ONLY 3 elements"
Small text → Regenerate with "150pt+" or "GIANT BOLD (150pt+)"
Cluttered → Regenerate with "60% white space" and "ULTRA-SIMPLE"
Complex workflow → SPLIT into multiple simple 3-element graphics

Visual Element Guidelines

⚠️ CRITICAL: Each graphic must have ONE message and MAXIMUM 3-4 elements.

ABSOLUTE LIMITS - These are NOT guidelines, these are HARD LIMITS:

MAXIMUM 3-4 elements per graphic (3 is ideal)
MAXIMUM 10 words total per graphic
MINIMUM 50% white space (60% is better)
MINIMUM 120pt for key numbers/metrics
MINIMUM 80pt for labels

For each poster section - STRICT requirements:

Section	Max Elements	Max Words	Example Prompt (REQUIRED PATTERN)
Introduction	3 icons	6 words	"POSTER FORMAT for A0: ULTRA-SIMPLE 3 icons: [icon1] [icon2] [icon3]. ONE WORD labels (100pt bold). 60% white space. 3 words total."
Methods	3 boxes	6 words	"POSTER FORMAT for A0: ULTRA-SIMPLE 3-box workflow: 'STEP1' → 'STEP2' → 'STEP3'. GIANT labels (120pt+). 60% white space. 3 words only."
Results	2-3 bars	6 words	"POSTER FORMAT for A0: TWO bars: 'BASELINE 70%' 'OURS 95%'. GIANT percentages (150pt+) ON bars. NO axis. 60% white space."
Conclusions	3 cards	9 words	"POSTER FORMAT for A0: THREE cards: '95%' (150pt) 'ACCURATE', '2X' (150pt) 'FASTER', checkmark 'READY'. 60% white space."
Case Study	3 elements	5 words	"POSTER FORMAT for A0: ONE case: logo + '18 MONTHS' (150pt) + 'to discovery' (60pt). 60% white space."
Timeline	3 points	3 words	"POSTER FORMAT for A0: THREE years only: '2018' '2021' '2024' (150pt each). Large icons. 60% white space. NO details."

MANDATORY prompt elements (ALL required, NO exceptions):

"POSTER FORMAT for A0" - MUST be first
"ULTRA-SIMPLE" or "ONLY X elements" - content limit
"GIANT (120pt+)" or specific font sizes - readability
"60% white space" - mandatory breathing room
"readable from 10-12 feet" - viewing distance
Exact count of words/elements - "3 words total" or "ONLY 3 icons"

PATTERNS THAT ALWAYS FAIL (REJECT IMMEDIATELY):

❌ "7-stage drug discovery workflow" → Split to "3 mega-stages"
❌ "Timeline from 2015-2024 with annual updates" → "ONLY 3 key years"
❌ "3 case studies with details" → Make 3 separate simple graphics
❌ "Comparison of 5 methods with metrics" → "ONLY 2: ours vs best"
❌ "Complete architecture showing all layers" → "3 components only"
❌ "Show stages 1,2,3,4,5,6" → "3 high-level stages"

PATTERNS THAT WORK:

✅ "3 mega-stages collapsed from 7" → Proper simplification
✅ "ONE case with ONE metric" → Will make multiple if needed
✅ "ONLY 3 milestones" → Selective, focused
✅ "2 bars: ours vs baseline" → Direct comparison
✅ "3-component high-level view" → Appropriately simplified