Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Results Analysis for ML/AI Research · OpenLAIR | DecimalAI

Inno Experiment Analysis

SkillScore22Maintained

inno-experiment-analysis

This skill should be used when the user asks to "analyze experimental results", "generate results section", "statistical analysis of experiments", "compare model performance", "create results visualization", or mentions connecting experimental data to paper writing. Provides comprehensive guidance for analyzing ML/AI experimental results and generating paper-ready content.

AI/MLby OpenLAIR

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Research packResearchAnalysisStatistics

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/inno-experiment-analysis/SKILL.md

decimalai skills pull inno-experiment-analysis --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/inno-experiment-analysis/SKILL.md https://app.decimal.ai/s/inno-experiment-analysis/SKILL.md

Raw file: /s/inno-experiment-analysis/SKILL.md ↗ · machine-readable summary at /s/inno-experiment-analysis.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

23 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 56d ago

Yes — +13% lift · at higher cost+59% tokens0% turns22% of whole cases pass with skill (5/23)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-02	✗→✓	▲ Improved	—	—
case-22	✗→✓	▲ Improved	—	—
case-01	✗→✓	▲ Improved	—	—
case-21	✓→✓	＝ Same ✓	—	—
case-07	✓→✓	＝ Same ✓	—	—

Showing 5 of 23 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 7/30/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

OpenLAIR / dr-claw

skills/inno-experiment-analysis/SKILL.md

View on GitHub ↗

License NOASSERTIONImported 8/2/2026Version v1Bundled 4 files

Skill contents

references/

SKILL.md

Results Analysis for ML/AI Research

A systematic experimental results analysis workflow connecting experimental data to paper writing.

Core Features

This skill provides three core capabilities:

Experimental Data Analysis - Read and analyze experimental data in various formats
Statistical Validation - Perform statistical significance tests and performance comparisons
Paper Content Generation - Generate text and visualizations for the Results section

When to Use

Use this skill when you need to:

Analyze experimental results (CSV, JSON, TensorBoard logs)
Generate the Results section of a paper
Compare performance across multiple models
Perform statistical significance tests
Create publication-quality visualizations
Validate the reliability of experimental results

Workflow

Standard Analysis Pipeline

Data Loading → Data Validation → Statistical Analysis → Visualization → Writing → Quality Check

Step 1: Data Loading and Validation

Supported Data Formats:

CSV files - Tabular data
JSON files - Structured results
TensorBoard logs - Training curves
Python pickle - Complex objects

Data Validation Checks:

Completeness check - Missing values, outliers
Consistency check - Data format, units
Reproducibility check - Random seeds, version info

Select appropriate tools for data loading and preliminary validation based on data format.

Step 2: Statistical Analysis

Basic Statistics:

Mean
Standard Deviation
Standard Error
Confidence Interval

Significance Tests:

t-test - Two-group comparison
ANOVA - Multi-group comparison
Wilcoxon test - Non-parametric test
Bonferroni correction - Multiple comparison correction

Select appropriate statistical tests based on data characteristics.

Key Principles:

Report complete statistical information (mean ± std/SE)
Specify the test method and significance level used
Report p-values and effect sizes
Consider multiple comparison issues

See references/statistical-methods.md for the complete statistical methods guide.

Step 3: Model Performance Comparison

Comparison Dimensions:

Accuracy/Performance metrics
Training time/Inference speed
Model complexity/Parameter count
Robustness/Generalization ability

Comparison Methods:

Baseline comparison - Compare with existing methods
Ablation study - Validate component contributions
Cross-dataset validation - Test generalization

Systematically compare performance across different methods, ensuring fair comparison.

Step 4: Visualization

Publication-Quality Visualization Requirements:

Vector format (PDF/EPS)
Colorblind-friendly palette
Clear labels and legends
Appropriate error bars
Readable in black-and-white print

Common Chart Types:

Line chart - Training curves, trend analysis
Bar chart - Performance comparison
Box plot - Distribution display
Heatmap - Correlation analysis
Scatter plot - Relationship display

Use appropriate visualization tools to generate publication-quality figures.

See references/visualization-best-practices.md for the visualization guide.

Step 5: Writing the Results Section

Results Section Structure:

markdown
## Results

### Overview of Main Findings
[1-2 paragraphs summarizing core results]

### Experimental Setup
[Brief description of experimental configuration; details in appendix]

### Performance Comparison
[Comparison with baseline methods, including tables and figures]

### Ablation Study
[Validate contributions of each component]

### Statistical Significance
[Report statistical test results]

### Qualitative Analysis
[Case studies, visualization examples]

Writing Principles:

Clearly state the hypothesis each experiment validates
Guide readers to observe key phenomena: "Figure X shows..."
Report complete statistical information
Honestly report limitations

See references/results-writing-guide.md for the complete writing guide.

Step 6: Quality Check

Checklist:

] All values include error bars/confidence intervals
] Statistical test methods are specified
] Figures are clear and readable (including black-and-white print)
] Hyperparameter search ranges are reported
] Computational resources are specified (GPU type, time)
] Random seed settings are specified
] Results are reproducible (code/data available)

Common Mistakes and Pitfalls

Statistical Errors

❌ Wrong approach:

Reporting only the best results (cherry-picking)
Confusing standard deviation and standard error
Not reporting statistical significance
Not correcting for multiple comparisons

✅ Correct approach:

Report all experimental results
Clearly specify whether standard deviation or standard error is used
Perform appropriate statistical tests
Use Bonferroni or similar correction methods

Visualization Errors

❌ Wrong approach:

Using non-colorblind-friendly palettes
Y-axis not starting from 0 (exaggerating differences)
Missing error bars
Overly complex figures

✅ Correct approach:

Use Okabe-Ito or Paul Tol palettes
Set reasonable axis ranges
Include error bars and confidence intervals
Keep figures clean and clear

Writing Errors

❌ Wrong approach:

Over-interpreting results
Not describing experimental setup
Hiding negative results
Missing statistical information

✅ Correct approach:

Objectively describe observed phenomena
Provide sufficient experimental details
Honestly report all results
Report complete statistical information

See references/common-pitfalls.md for the complete error patterns and fixes.

Integration with Paper Writing

Collaboration with ml-paper-writing Skill

This skill focuses on experimental results analysis and works in tandem with the ml-paper-writing skill:

inno-experiment-analysis handles:

Data analysis and statistical tests
Visualization generation
Results interpretation

ml-paper-writing handles:

Complete paper structure
Citation management
Conference format requirements

Workflow Integration:

Experiments complete → inno-experiment-analysis analyzes
    ↓
Generate analysis report and visualizations
    ↓
ml-paper-writing integrates into paper
    ↓
Complete Results section

Output Format

After analysis, the following are generated:

Analysis Report (analysis-report.md)

Statistical summary
Key findings
Suggested figures

Visualization Files (figures/)

PDF format figures
Standalone figure captions

Results Draft (results-draft.md)

Text ready for direct use in the paper
Includes figure references

Examples and Templates

Example Files

Refer to the examples/ directory for complete examples:

example-analysis-report.md - Complete analysis report example
example-results-section.md - Paper Results section example

Workflow Overview

The complete analysis pipeline includes:

Data Loading - Read results from experiment output files
Statistical Analysis - Compute basic statistics and perform significance tests
Visualization - Create publication-quality figures
Report Generation - Integrate analysis results and visualizations

See the guides in the references/ directory for detailed methods and best practices.

Reference Resources

Detailed Guides

references/statistical-methods.md - Complete statistical methods guide
references/results-writing-guide.md - Results section writing standards
references/visualization-best-practices.md - Visualization best practices
references/common-pitfalls.md - Common errors and fixes

External Resources

Best Practices Summary

Data Analysis

✅ Recommended:

Run experiments multiple times (at least 3-5 runs)
Report complete statistical information
Use appropriate statistical tests
Check data completeness

❌ Prohibited:

Cherry-picking best results
Ignoring statistical significance
Hiding negative results
Not reporting experimental setup

Visualization

✅ Recommended:

Use vector format
Colorblind-friendly palettes
Include error bars
Clear labels

❌ Prohibited:

Raster formats (PNG/JPG)
Misleading axis scales
Overly complex figures
Missing legends

Writing

✅ Recommended:

Objectively describe results
Provide sufficient detail
Honestly report limitations
Guide reader attention

❌ Prohibited:

Over-interpretation
Hiding details
Exaggerating effects
Vague descriptions

Summary

This skill provides a systematic experimental results analysis workflow:

Data Loading and Validation - Ensure data quality
Statistical Analysis - Perform appropriate statistical tests
Model Comparison - Systematic performance comparison
Visualization - Publication-quality figures
Writing - Results section content
Quality Check - Ensure reproducibility

Following these principles produces high-quality, reproducible experimental results analysis that meets top conference standards.

Illustrative Want to see what inno-experiment-analysis would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

7/27/2026, 15:03 UTC · 23 cases

Lift

+13%

22% vs 9%

Agent turns

1.0 vs 1.0

Tokens

+59%

4,351.26 vs 2,740.17

fail → pass · 3pass → pass · 2fail → fail · 18

Per-case results (23)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-10 We added three novel components (A, B, and C) to a baseline neural network. To prove that all three components work together, we ran two experiments: baseline alone vs baseline + A + B + C. Is this comparison sufficient for an ablation study in the Results section?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-03 Can you take our training logs and CSV files from TensorBoard and generate a full experimental evaluation package? Specifically, we want a comprehensive analysis report with statistical metrics, vector format plots showing baseline comparisons with error bars, a drafted Results section including ablation analysis, and a final quality checklist audit ensuring all reproducibility standards are met.	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-08 We collected results from three independent researchers who trained our model on different workstations. Before combining their outputs into a single benchmark table, what specific validation checks should be performed on the raw files?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-02 We finished our multi-seed ablation experiments and saved the raw JSON logs containing accuracy and inference time metrics. I need you to analyze these results, perform statistical significance testing, and produce three deliverable files: a statistical analysis report, publication-quality vector charts for the performance comparison, and a structured markdown results draft suitable for inclusion in a machine learning paper.	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-04 We are comparing our proposed model against 5 baseline models on a classification benchmark across 10 random seeds. Should we run a series of unadjusted independent two-sample t-tests between our model and each baseline to declare statistical superiority at p < 0.05?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-20 We evaluated our object detection algorithm across 5 random seeds on the COCO dataset and got strong mean precision metrics. Is evaluating on COCO alone sufficient to demonstrate overall model capability across all comparison dimensions?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-06 We are building bar charts and training curve line plots for our ML conference submission. A teammate suggested using the standard 'jet' or 'rainbow' color scale because it renders vivid high-contrast colors across 6 model variations. Should we use rainbow/jet colors for vector figures?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-19 We are reporting error bounds on our model accuracy in a results table where sample size n=10 seeds. A reviewer asks whether our reported ± bounds represent sample variability or precision of the estimated mean. Which metric should be reported for each case?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-21 We loaded a CSV dataset of training runs across 20 hyperparameter trials. Some trials terminated early with NaN loss values, and two trials logged inference latencies 1000x higher than normal due to disk paging. How should these issues be categorized and handled during data validation?	pass→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-23 We are writing up the Statistical Significance subsection for our results section. We plan to state 'Our model was significantly better than baseline (p < 0.05)' without mentioning test names, effect sizes, or exact p-values. Does this meet statistical reporting standards?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-17 Here is our draft text. Please wrap it in the official NeurIPS 2025 LaTeX style file format (`neurips_2025.sty`), set up the page geometry, setup custom math macro packages, and fix alignment errors in the preamble.	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-07 Our new model achieves 94.2% accuracy while the best baseline gets 93.8%. On a chart with a 0% to 100% Y-axis, the bars look almost identical. To highlight our victory in the paper's main chart, we plan to set the Y-axis range strictly from 93.5% to 94.5%. Is this recommended for publication-quality figures?	pass→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-22 When completing an experimental results analysis pipeline on raw run metrics, what are the exact three primary output artifacts that should be generated for integration into a research paper?	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-05 We evaluated two language model prompts on a benchmark where human preference ratings are ordinal scores from 1 to 5 with a highly skewed non-normal distribution. Someone suggested using a standard parametric two-sample t-test to check if the mean score difference is statistically significant. Is a t-test appropriate here?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-13 We created colored line plots for 4 model variations using distinct shades of green and blue. Is checking that colors look distinct on a bright LED monitor sufficient for quality validation of paper figures?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-01 I have raw evaluation metrics stored in a CSV file comparing our proposed AI model against four baseline architectures across five random seeds. Please process this dataset, run appropriate statistical validation tests comparing performance and speed, and draft the complete markdown Results section for our research paper. Make sure to generate the corresponding vector figure files with error indicators and produce an analysis summary report outlining key findings.	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-15 Can you write the complete PyTorch implementation of a 4-layer Transformer encoder module, including multi-head attention, layer normalization, residual connections, and forward pass methods?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-11 We are organizing the markdown structure for our paper's Results section. We already have sections for Overview of Main Findings, Experimental Setup, Performance Comparison, Ablation Study, and Statistical Significance. What remaining section should be included according to standard ML results section structure?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-16 Please write the complete Introduction section and Related Work section for our paper on vision transformers. Include a literature survey of self-attention mechanisms from 2017 to 2024 and outline our key motivations.	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-09 In our benchmark table, we plan to display only the single highest accuracy score achieved across 10 random seeds for each algorithm to show what each model is capable of at its peak performance. Is this acceptable for top-tier conference paper results?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-14 I need to clean up my BibTeX reference file for an arXiv paper submission. Can you scan these 15 BibTeX entries, fix missing venue names, format author lists into standard BibTeX conventions, and resolve duplicate citation keys?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-12 We are performing a final quality check audit on our experimental results documentation prior to submitting to a machine learning conference. Beyond metrics, statistical tests, and error bars, what hardware and compute details must be explicitly documented in the quality check list?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-18 When saving generated visualizations for our experimental results analysis, our script defaults to exporting high-resolution 300 DPI PNG images. Should we keep PNG as the primary format for the publication figure files?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +13 percentage points is the difference between those two pass rates over the 23 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 8.7%
Pass rate with the skill: 21.7%
Cases attempted: 23
Cases counted toward lift: 23
Cases that regressed: 0
Skill version measured: v1
Run completed: 7/27/2026, 15:03 UTC

The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run22

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+64% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 56 days ago

Other measured skills in the registry, with their headline benchmark lift.

Ready to supercharge your AI agents?

Inno Experiment Analysis

SkillSafety

Source

Results Analysis for ML/AI Research

Core Features

When to Use

Workflow

Standard Analysis Pipeline

Step 1: Data Loading and Validation

Step 2: Statistical Analysis

Step 3: Model Performance Comparison

Step 4: Visualization

Step 5: Writing the Results Section

Step 6: Quality Check

Common Mistakes and Pitfalls

Statistical Errors

Visualization Errors

Writing Errors

Integration with Paper Writing

Collaboration with ml-paper-writing Skill

Output Format

Examples and Templates

Example Files

Workflow Overview

Reference Resources

Detailed Guides

External Resources

Best Practices Summary

Data Analysis

Visualization

Writing

Summary

How this skill was measured

Related skills