lipu-jpg Paper Hypothesis
Generate and validate research hypotheses based on identified research gaps.
Hypothesis Generation Framework
1. Gap Analysis
Based on literature review output, identify:
- What problem remains unsolved
- Why current approaches fail
- What new angle can be explored
2. Hypothesis Structure
H1: [Independent variable] affects [dependent variable] because [mechanism]
Example:
H1: Using larger context windows improves reasoning performance
in LLM benchmarks because longer contexts provide more
contextual information for inference.
3. Feasibility Assessment
- Data availability: Can you obtain necessary data?
- Method feasibility: Can you implement the approach?
- Resource requirements: Time, compute, tools needed
- Validation path: How will you prove/disprove the hypothesis?
Experiment Design
| Component | Description |
|---|---|
| Control | Baseline method to compare against |
| Treatment | Your proposed approach |
| Metrics | How to measure success |
| Statistics | Test selection, significance level |
Output Format
{
"hypotheses": [
{
"id": "H1",
"statement": "...",
"variables": {"independent": "...", "dependent": "..."},
"feasibility": {"data": "available", "method": "feasible"},
"experiment_design": {...}
}
],
"recommended_experiments": ["exp1", "exp2"],
"potential_challenges": ["challenge1", "challenge2"]
}
Tips
- Start with one clear hypothesis, add secondary ones later
- Make hypotheses falsifiable (provable wrong)
- Consider both positive and negative results as publishable
- Use brainstorming skill for idea exploration |- Use brainstorming skill for idea exploration
Scripts
scripts/hypothesis-generator.py: Generate hypotheses, assess feasibility, design experiments