Neural reranking

신경망 재순위화

Rescore first-stage candidates together with the query and reorder them.

···
html
<div class="v"><header><b>RERANK</b><span id="state"></span></header><main><div class="rows" id="rows"></div></main><footer>candidate order → model order</footer></div>
css
.v{width:min(92vw,660px);height:min(86vh,310px);box-sizing:border-box;padding:clamp(9px,2.5vmin,18px);border:1px solid var(--line);border-radius:14px;background:var(--surface);display:flex;flex-direction:column;gap:7px;color:var(--fg);font:500 clamp(15px,4.5vmin,19px)/1.25 var(--font-sans, sans-serif)}.v header,.v footer{display:flex;justify-content:space-between;align-items:center;gap:8px;white-space:nowrap}.v header b{color:var(--accent);font-size:.9em}.v header span,.v footer{color:var(--muted);font-size:.82em}.v main{flex:1;min-height:0;position:relative;overflow:hidden}.v .mono{font-family:ui-monospace,SFMono-Regular,Consolas,monospace}.v .active{background:var(--accent)!important;color:var(--bg)!important;border-color:var(--accent)!important}.v .muted{opacity:.45} .rows{height:100%;display:flex;flex-direction:column;justify-content:center;gap:5px}.row{height:25%;border:1px solid var(--line);border-radius:5px;padding:4px 8px;display:flex;align-items:center;justify-content:space-between;transition:transform .35s}.row b{font:600 1em ui-monospace,monospace}.row span{color:var(--accent);font:600 .85em ui-monospace,monospace}
js
const orders=[["A","B","C"],["B","C","A"]],scores={A:"0.42",B:"0.91",C:"0.73"};let n=0;function draw(){document.getElementById("rows").innerHTML=orders[n].map((x,i)=>"<div class=\"row "+(i===0?"active":"")+"\"><b>"+(i+1)+"  DOC "+x+"</b><span>"+(n?scores[x]:"candidate")+"</span></div>").join("");document.getElementById("state").textContent=n?"model score":"first pass";n=1-n}draw();setInterval(draw,1400)

A fast first-stage search gathers candidates. A reranking model then reads each query-document pair and assigns a more precise relevance score.

Applying the model to every document would be expensive, so reranking uses a limited candidate set. It cannot recover a relevant document missed by the first stage.

When to use

Use it when top result ordering materially affects answer quality.

Open as page ↗