Practical notes: Reranking for RAG: Cross-Encoders, LLM Rerankers, and Latency

The following notes reconstruct a practical path around “Reranking for RAG: Cross-Encoders, LLM Rerankers, and Latency Tradeoffs”. Emphasis stays on contracts, checks, and drop-in code placeholders rather than motivational framing. When working through Overview, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

The Bridge from Retrieval to Ranking

The Bridge from Retrieval to Ranking works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices.

What Reranking Actually Does

What Reranking Actually Does works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices.

Why First-Pass Retrieval is Noisy by Design

Why First-Pass Retrieval is Noisy by Design works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices. Why First-Pass Retrieval is Noisy by Design works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

The Two Main Reranking Families

For The Two Main Reranking Families, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Cross-Encoders are the Practical Default

For Cross-Encoders are the Practical Default, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

import cohere
import time

def rerank_cross_encoder(
    query: str,
    candidates: list[dict],
    top_n: int = 5,
    model: str = "rerank-v4.0-pro",
) -> list[dict]:
    """
    The practical default for second-stage ranking.
    Passes the query and candidate texts to a dedicated cross-encoder model.
    """
    co = cohere.ClientV2()

    # Extract just the text content for the API call
    documents = [c["content"] for c in candidates]

    resp = co.rerank(
        model=model,
        query=query,
        documents=documents,
        top_n=top_n,
    )

    # Reattach the original metadata and the new score
    reranked = []
    for r in resp.results:
        original_chunk = candidates[r.index]
        reranked.append({
            **original_chunk,
            "rerank_score": r.relevance_score
        })

    return reranked

LLM Rerankers are Flexible and Expensive

For LLM Rerankers are Flexible and Expensive, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call. For LLM Rerankers are Flexible and Expensive, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

import anthropic

JUDGE_PROMPT = """\
You are a strict relevance judge. Given a user query and a candidate document chunk,
rate how well the chunk answers the query on a scale of 0 to 10.

Respond with ONLY a JSON object in this exact format:
{"score": <int>, "reason": "<one short sentence>"}

Query: {query}
Chunk: {chunk}"""

async def rerank_llm(
    query: str,
    candidates: list[dict],
    top_n: int = 5,
) -> list[dict]:
    """
    Expensive special forces. Uses an LLM to reason about nuance and completeness.
    """
    client = anthropic.AsyncAnthropic()
    scored = []

    for c in candidates:
        resp = await client.messages.create(
            model="claude-opus-4-6",
            max_tokens=128,
            messages=[{
                "role": "user",
                "content": JUDGE_PROMPT.format(query=query, chunk=c["content"]),
            }],
        )

        import json
        try:
            result = json.loads(resp.content[0].text)
            scored.append({
                **c,
                "rerank_score": result["score"],
                "reason": result.get("reason", "")
            })
        except (json.JSONDecodeError, KeyError):
            # Fallback if the model fails to follow JSON instructions
            scored.append({**c, "rerank_score": 0, "reason": "parse_error"})

    # Sort by the LLM-assigned score descending
    scored.sort(key=lambda x: x["rerank_score"], reverse=True)
    return scored[:top_n]

The Latency Tradeoff

When working through The Latency Tradeoff, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Cache stable system instructions and tool schemas. Re-sending identical preamble is a common source of burn.

def rerank_with_timing(
    rerank_fn: callable,
    query: str,
    candidates: list[dict],
    top_n: int = 5,
) -> tuple[list[dict], float]:
    """
    Measure the exact cost of the reranking stage.
    """
    t0 = time.perf_counter()

    results = rerank_fn(query, candidates, top_n)

    latency_ms = (time.perf_counter() - t0) * 1000
    return results, latency_ms

When Reranking is Worth It

When working through When Reranking is Worth It, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Cache stable system instructions and tool schemas. Re-sending identical preamble is a common source of burn.

When Reranking is Overkill

When working through When Reranking is Overkill, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Cache stable system instructions and tool schemas. Re-sending identical preamble is a common source of burn. When working through When Reranking is Overkill, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Reranking Failure Modes

Reranking Failure Modes works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices.

def dedupe_candidates(
    candidates: list[dict],
    similarity_threshold: float = 0.85,
) -> list[dict]:

    seen_tokens: list[set[str]] = []
    deduped = []

    for c in candidates:
        tokens = set(c["content"].lower().split())
        is_dup = False

        for s in seen_tokens:
            # Calculate simple Jaccard similarity
            overlap = len(tokens & s) / max(len(tokens | s), 1)
            if overlap >= similarity_threshold:
                is_dup = True
                break

        if not is_dup:
            deduped.append(c)
            seen_tokens.append(tokens)

    return deduped

from datetime import datetime, timezone

def apply_metadata_boost(
    candidates: list[dict],
    freshness_halflife_days: int = 90,
) -> list[dict]:

    now = datetime.now(timezone.utc)
    boosted = []

    for c in candidates:
        score = c.get("rerank_score", 0.0)

        # Hard penalty for superseded documentation
        if c.get("status") == "superseded":
            score *= 0.4

        # Gradual decay for older documents
        updated = c.get("updated_at")
        if updated:
            age_days = (now - updated).days
            decay_factor = max(0.5, 1 - age_days / (freshness_halflife_days * 2))
            score *= decay_factor

        boosted.append({**c, "rerank_score": score})

    # Sort again based on the adjusted scores
    boosted.sort(key=lambda x: x["rerank_score"], reverse=True)
    return boosted

How to Evaluate Reranking Properly

How to Evaluate Reranking Properly works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices.

from dataclasses import dataclass

@dataclass
class RerankEvalCase:
    query: str
    expected_substring: str
    category: str  # e.g., "identifier", "procedure", "troubleshooting"

def eval_reranking(
    cases: list[RerankEvalCase],
    retrieve_fn: callable,
    rerank_fns: dict[str, callable | None],
    top_n: int = 3,
) -> dict:
    """
    Compare multiple reranking strategies against a baseline.
    Measures hit rate at top-N and tracks latency overhead.
    """
    results = {}

    for name, rerank_fn in rerank_fns.items():
        hits = 0
        total_latency = 0.0
        by_type: dict[str, dict] = {}

        for case in cases:
            # Get the exact same starting candidates for every strategy
            candidates = retrieve_fn(case.query)

            if rerank_fn is not None:
                reranked, lat = rerank_with_timing(
                    rerank_fn, case.query, candidates, top_n
                )
                total_latency += lat
            else:
                # Baseline: just take the top-N from first-pass retrieval
                reranked = candidates[:top_n]

            # Check if the expected evidence made it into the final prompt window
            top_contents = [r["content"] for r in reranked]
            found = any(case.expected_substring in c for c in top_contents)
            hits += int(found)

            # Track metrics by query category
            by_type.setdefault(case.category, {"hit": 0, "total": 0})
            by_type[case.category]["total"] += 1
            by_type[case.category]["hit"] += int(found)

        total = len(cases)
        results[name] = {
            "hit_rate": hits / total if total else 0,
            "avg_latency_ms": total_latency / total if total else 0,
            "by_type": {
                t: {**v, "rate": v["hit"] / v["total"]}
                for t, v in by_type.items()
            },
        }

    return results

A Practical Default Recommendation

A Practical Default Recommendation works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices. A Practical Default Recommendation works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

def two_stage_retrieve(
    query: str,
    retrieve_fn: callable,
    top_k: int = 20,
    top_n: int = 5,
) -> tuple[list[dict], dict]:
    """
    The complete production pipeline for second-stage ranking.
    """
    t0 = time.perf_counter()

    # Stage 1: Fast hybrid retrieval
    candidates = retrieve_fn(query)[:top_k]

    # Clean up the candidate pool
    candidates = dedupe_candidates(candidates)

    # Stage 2: Cross-encoder rerank
    # We score slightly more than top_n to allow metadata boosts to reorder the edges
    score_limit = min(top_n * 2, len(candidates))
    reranked = rerank_cross_encoder(query, candidates, top_n=score_limit)

    # Apply business logic for freshness and status
    final = apply_metadata_boost(reranked)[:top_n]

    latency = (time.perf_counter() - t0) * 1000

    trace = {
        "query": query,
        "first_pass_count": len(candidates),
        "post_rerank_count": len(reranked),
        "final_count": len(final),
        "latency_ms": latency,
    }

    return final, trace

What’s Next

For What’s Next, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Continue Reading

For Continue Reading, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Operational checklist

For Operational checklist, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state.

Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Measure recall on a fixed question set before tuning prompts. Prompt churn rarely fixes a weak retrieval surface.

Pin dependency versions and record the image digest that ran the demo. Reproducibility beats tribal knowledge.

Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline.

Before promoting the stack, freeze versions, capture a golden transcript for the critical path, and confirm rollback steps. Shared environments need rate limits, tenancy checks, and a clear owner for secret rotation. Prefer boring reliability over clever one-off demos.

Batch note for cdeb69942ea2: keep provider keys out of the repo, set a per-session token ceiling, and store transcripts next to the eval fixtures so later model swaps stay comparable.

When working through hardening note 0, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline.

Hardening detail 0/916: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

hardening note 1 works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Hardening detail 1/916: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

For hardening note 2, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish.

Hardening detail 2/916: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

When working through hardening note 3, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

Hardening detail 3/916: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

hardening note 4 works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Hardening detail 4/916: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.