Xiangdong-415 STEM Result Analyzer
Purpose
Use this skill to turn raw results into clear scientific claims. It helps decide what the results actually support, how to present figures and tables, and how to write a credible discussion.
Inputs To Ask For
- Result tables, figures, logs, summary metrics, or raw data files.
- Meaning of each metric and whether higher or lower is better.
- Baselines and experiment settings.
- Main claims the paper wants to support.
- Any surprising, negative, or inconsistent findings.
Workflow
Verify result context.
- Confirm datasets, sample sizes, units, seeds, repeats, and metrics.
- Check whether comparisons are apples-to-apples.
Extract primary findings.
- State the largest supported finding first.
- Separate statistically meaningful changes from noise.
- Identify trade-offs, such as accuracy versus latency.
Interpret ablations.
- Explain what each component contributes.
- Avoid claiming causality unless the design supports it.
Analyze failures.
- Identify cases where the method underperforms.
- Connect failures to assumptions, data distribution, physical constraints, or model limitations.
Build figure and table narrative.
- Each figure answers one question.
- Captions should state the experimental setting and main takeaway.
- Text should interpret, not repeat every number.
Write discussion and limitations.
- Explain why results occur.
- State practical significance.
- Name limitations honestly and turn them into future work.
Output Format
Provide:
- Key finding bullets.
- Figure and table captions or revised captions.
- Results-section draft.
- Discussion-section draft.
- Limitations paragraph.
- Reviewer-risk notes for weak, missing, or ambiguous evidence.
Quality Checklist
- Claims are no stronger than the data.
- Important negative results are not hidden.
- Uncertainty, variability, or statistical support is reported when needed.
- Figures and tables have a clear purpose.
- The discussion explains mechanisms and implications, not just rankings.