# Paperlab Hypothesis To Benchmark Matrix

> Convert PaperLab research questions into falsifiable hypotheses, benchmark matrices, statistical tests, and acceptance criteria for computational papers.

- Skill: `equinor/paperlab-hypothesis-to-benchmark-matrix` (Agent Skill)
- Install (CLI): `npx skillmds@latest add equinor/paperlab-hypothesis-to-benchmark-matrix`
- Raw SKILL.md: https://api.skillmd.com/api/skills/equinor/paperlab-hypothesis-to-benchmark-matrix/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: equinor (https://skillmd.com/u/equinor)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/equinor/paperlab-hypothesis-to-benchmark-matrix

---


# PaperLab Hypothesis to Benchmark Matrix

## When to Use

USE WHEN: a paper idea needs to become an executable benchmark plan before
`benchmark-runner` starts work.

## Required Elements

Every hypothesis must define:

- claim wording,
- null expectation or baseline,
- controlled variables,
- varied variables,
- response metric,
- minimum case count,
- statistical test or deterministic acceptance criterion,
- expected failure or boundary modes,
- source of reference values when relevant.

## Benchmark Matrix Pattern

```json
{
  "hypotheses": [
    {
      "id": "H1",
      "claim": "candidate solver reduces failures for near-critical mixtures",
      "metric": "failure_rate_pct",
      "baseline": "existing solver",
      "test": "paired proportion test",
      "acceptance": "p < 0.05 and absolute reduction >= 5 percentage points",
      "case_categories": ["near-critical", "two-phase", "trace-water"]
    }
  ]
}
```

## Pass Criteria

- No claim exists without a metric.
- Boundary cases are explicit, not accidental.
- Benchmark cases can be regenerated from configuration.
- The matrix is small enough to rerun during revision.

## Safety Rules

- Do not retrofit metrics after seeing results unless clearly marked exploratory.
- Do not mix validation data and training/tuning data without disclosure.
- Do not use proprietary benchmark inputs in public capsules unless cleared.