abstract-writer

Abstract Writer

Purpose

Draft or revise an abstract that makes a clear, evidence-backed promise about the paper. The abstract should help reviewers understand the problem, gap, method, result, and implication without exaggerating unsupported claims.

When To Use

• Creating the first abstract from a paper context.
• Revising an abstract after experiments change.
• Compressing an abstract for venue constraints.
• Checking whether an abstract overclaims novelty or results.

Inputs

Required:

• target venue and paper type;
• one-sentence thesis;
• main contributions;
• strongest verified result;
• task, dataset, metric, or validation protocol when central;
• known limitations or scope constraints.

Optional:

• current abstract;
• target word limit;
• novelty-scout report;
• claim-evidence matrix.

Abstract Structure

Use this five-move structure unless the venue strongly suggests otherwise:

1. Problem: what concrete problem matters?
2. Gap: what is missing or insufficient in prior work?
3. Method/insight: what does this paper introduce?
4. Evidence: what result supports the claim?
5. Implication: what changes, within the proven scope?

Procedure

1. Extract all claims from the current abstract or paper context.
2. Mark each claim as problem, gap, method, result, implication, or unsupported.
3. Require at least one concrete evidence sentence if the paper has experimental or empirical claims.
4. Replace vague adjectives with measurable scope: "improves Dice by X", "reduces memory by Y", "validated on Z", or "under assumption A".
5. Remove or weaken "first", "novel", "state-of-the-art", "robust", "clinical-grade", and similar terms unless backed by evidence.
6. Ensure the abstract does not introduce claims absent from the paper body.
7. Produce a revised abstract plus a claim-risk note.

Rubric

Dimension	Strong	Weak
Problem	Specific task/context	Generic field opening
Gap	Fair limitation in prior work	Strawman or vague "challenging" claim
Method	Clear mechanism or insight	Name-only method description
Evidence	Concrete result/validation	No numbers, datasets, or support
Scope	Honest implication	Broad conclusion beyond evidence
Readability	One coherent paragraph	List of disconnected claims