Experiment Cycle

Step 3: Full Benchmark (/bench-run)

• Run comprehensive benchmark suite based on PoC results
• Compare against relevant baselines from .research/methods.json
• Present results for user review at each stage

Step 4: Visualization (/result-viz)

• Generate standard comparison charts
• Create publication-quality figures
• User reviews and approves figures

Step 5: Limitation Analysis (/limitation)

• Analyze failure cases and boundary conditions
• Identify improvements for next iteration

Step 6: Cycle Summary

## Experiment Cycle Summary: [Idea]

### Hypothesis: [statement]
### Verdict: [supported / partial / refuted]

### Key Results
| Metric | Our Method | Best Baseline | Improvement |
|--------|-----------|---------------|-------------|

### Strengths
- ...

### Limitations
- ...

### Next Steps
- [ ] Draft method section with /draft-section
- [ ] Run /ablation for component analysis
- [ ] Consider /review-check

Output

• Console: cycle summary with verdict
• Files: experiment scripts, result files, figures, updated .research/experiments.json