The best research gets rejected when framed poorly.
Framing is not spin — it's helping reviewers see what you see.
Framing answers the reviewer's first question before they ask it:
The core narrative: gap → bridge. Every strong paper tells this story. The gap is why you exist. The bridge is what you built.
What the field lacks. Be specific. "Existing KG-based RS methods ignore cold-start users" is a gap. "LLM-based RS has limitations" is not — it's too vague to position anything.
What you provide. Must connect directly to the gap. If your bridge doesn't address the gap you stated — reviewers notice. Your method section will feel disconnected from your introduction.
"Recommender systems suffer from data sparsity and cold-start problems, which are well-known challenges."
"KG-augmented methods improve warm users but degrade cold-start performance because popular items dominate KG coverage — a contradiction we resolve by..."
You define the problem differently. New objective function, new task definition, new constraint. Hardest to do but highest impact.
New model, algorithm, or architecture solving an existing problem. Most common in RS. Must outperform state-of-the-art on fair baselines.
A proof, bound, or analysis that explains why something works. Rare in RS but highly respected. Adds to M5 skills.
Rigorous experiments revealing something the field didn't know. "Method X, assumed to work because of Y, actually works because of Z."
Critical: Clearly state which type you're contributing at the start of your introduction. Reviewers who expect Type 2 but get Type 4 will reject the paper — not because it's bad but because the framing misled them.
Every contribution statement needs exactly three sentences:
"Existing [methods/approaches] fail to address [specific limitation] because [root cause]."
"We propose [method name], which [core idea] by [key mechanism]."
"Experiments on [datasets] demonstrate [specific improvement] over [strong baselines]."
Applied to a hypothetical RS paper:
"KG-augmented LLM recommenders retrieve structural knowledge but ignore whether an item has sufficient KG coverage — causing silent degradation for 95% of items in sparse KGs."
"We propose Coverage-Aware KG-RAG, which dynamically selects between text-only and graph-augmented generation based on per-item KG density."
"Experiments on MovieLens and Amazon Book show 12% improvement on items with sparse KG coverage with no degradation on dense items."
Every paper. Every time. The gap must be specific. The bridge must connect directly to it.
New formulation, method, theory, or empirical finding. State it explicitly. Reviewers who expect the wrong type reject good papers.
Gap → Bridge → Evidence. Write this first. If you can't write it clearly — you don't understand your own contribution yet.
Next: M6 · L2 — Writing a Strong Introduction