Walk through the parallel research plus reducer pattern and what the reducer's job actually is

Attempt it

The reducer's job is synthesis, not concatenation. Its inputs are N partial answers, possibly contradictory, possibly overlapping, possibly with varying confidence. Its output is one coherent answer that resolves conflicts, deduplicates redundant facts, attributes claims to sources, and surfaces remaining uncertainty.

Common failure modes when the reducer is weak. One, paste and bullet: the reducer concatenates worker outputs into a long list and calls it done; conflicts go unresolved and the reader has to do the synthesis. Two, lowest common denominator: the reducer keeps only facts that appear in multiple workers, losing the unique but correct findings each worker contributed. Three, halluciation on conflict: when workers disagree the reducer invents a compromise that none of them supports. The fix is a strong reducer prompt that names these failure modes, gives an explicit conflict-resolution rubric, and forces a source attribution per claim so the reader can verify.

Your answer

The **parallel research plus reducer pattern** fans out N research workers (each given a slightly different prompt or a different sub-question) and joins their outputs through a final reducer agent. The pattern is popular for research workflows because parallel workers reduce wall-clock time and explore the space more broadly than a single agent would. The reducer's actual job is **synthesis, not concatenation**. Its inputs are N partial answers that are possibly contradictory, possibly overlapping, possibly with varying confidence. Its output is one coherent answer that resolves conflicts, deduplicates redundant facts, attributes claims to sources, and surfaces remaining uncertainty. Three common failure modes happen when the reducer is weak. **Paste and bullet**: the reducer concatenates worker outputs into a long list and calls the result done. Conflicts go unresolved, the reader has to do the synthesis work the reducer was supposed to do, and the output is just longer than any individual worker's. **Lowest-common-denominator**: the reducer keeps only facts that appear in multiple workers. This sounds safe but it loses the unique but correct findings each worker contributed; the output becomes shorter than any individual worker's and misses the most valuable contributions. **Hallucinate-on-conflict**: when workers disagree, the reducer invents a compromise that none of them actually supports. This is the worst failure because it manufactures confident-sounding falsehoods. The fix is a strong reducer prompt: name the failure modes, give an explicit conflict-resolution rubric, and force source attribution per claim so the reader can verify.

The parallel research plus reducer pattern is one of the most common multi-agent topologies in research workflows, and the **reducer is where the pattern lives or dies**. The reducer's job is **synthesis**, not concatenation. It receives N partial answers from N workers and must produce one coherent answer. Synthesis means four things at once. **Conflict resolution**: when workers disagree, the reducer decides which to trust by source quality or surfaces the disagreement explicitly. **Deduplication**: when multiple workers report the same fact, report it once with consolidated source attribution. **Source attribution**: every load-bearing claim is tagged with the supporting source so the reader can verify. **Uncertainty surfacing**: facts no worker could verify are marked open rather than presented as confident. Three failure modes dominate when the reducer prompt is weak. **Paste and bullet**: the reducer concatenates worker outputs into a long list and exits. Conflicts go unresolved; the reader has to do the synthesis the reducer was supposed to do. **Lowest-common-denominator**: the reducer keeps only facts that appear in multiple workers. The output ends up shorter than any individual worker's, with the most valuable unique findings dropped. Root cause: conflating redundancy with reliability. **Hallucinate-on-conflict**: when workers disagree, the reducer invents a compromise none of them supports. Output contains confident-sounding facts no worker actually produced. The fix has three parts. A **strong reducer prompt** that names the failure modes and gives a conflict-resolution rubric (source-quality tier, recency, confidence). **Structured worker outputs** so the reducer has typed signal to synthesise. **Forced per-claim attribution** in the output schema so the final answer cannot include unsourced facts. The reducer prompt usually needs more iteration than the worker prompts because the reducer's task is harder.

Quick reference

TL;DR

The reducer's job is synthesis (resolve conflicts, dedupe, attribute, surface uncertainty), not concatenation; weak reducers fail as paste and bullet, lowest common denominator, or hallucinate on conflict.

Memory aid

Sign in for the mnemonic that makes this stick. See Pro pricing.

Easy to grasp

Imagine three friends each spend an hour researching the same restaurant. They come back with overlapping notes (all mention the menu), unique notes (one found the chef's old reviews, another tasted the desserts, the third read the wait-times data), and a couple of contradictions (one says it is quiet, another says it is loud). Now you sit down to write one single review for someone deciding whether to go. You do not staple the three friends' notes together and hand them over. You weigh what they said, resolve the noise versus quiet disagreement by noting it depends on the night, you pick the most-credible source on each fact, and you call out the things nobody actually verified. That writing the review step is the reducer's job in a parallel-research agent system, and it is much harder than just gluing the friends' notes together.

Key concepts

Multi-Agent Systems Agents Context Engineering

Detailed answer & concept explanation

~8 min read

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example. Click to expand.

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example.

Sign in for 5 free unlocks/day or go Pro for unlimited. See Pro pricing.

Quick scan · 2–4 min

Everything important, quickly.

State the synthesis not concatenation framing. Name the four parts of the reducer contract (resolve, dedupe, attribute, surface uncertainty). Walk the three failure modes and map each to the missing contract part. Close on structured worker outputs as the upstream fix and per-claim attribution as the output discipline.

Real-world examples

Real products, models, and research that use this idea.

OpenAI Deep Research uses a planner plus parallel workers plus synthesiser pattern with structured citation extraction in the synthesiser.
Perplexity Pro Search fans out multiple search workers and synthesises through a citation-attributing reducer prompt.
Anthropic's published agent-design guidance describes the same parallel plus reducer pattern with explicit conflict-resolution prompting.
LangGraph's Send primitive plus a synthesiser node is the canonical open-source implementation of this pattern.
CrewAI's hierarchical process uses a manager agent as the reducer that synthesises worker task outputs.

Interviewer follow-ups

What an interviewer would ask next. Try answering before peeking at the approach.

QWhy is lowest common denominator reduction a worse failure mode than paste and bullet, even though it produces a shorter output?

Sign in or go Pro to see the approach for this follow-up.

QHow would you evaluate a reducer regression-style so the three failure modes do not slip in across versions?

Sign in or go Pro to see the approach for this follow-up.

QWhat changes about the reducer's job when the upstream workers themselves are LLM-as-judge agents rather than information-gathering agents?

Sign in or go Pro to see the approach for this follow-up.

Don't say this

Red flags and common mistakes that signal junior thinking. Click to expand.

Most common mistake

Treating the reducer as a glue step that concatenates worker outputs, when the actual job is conflict resolution, deduplication, source attribution, and surfacing remaining uncertainty.

Sign in or go Pro to see all red flags and common mistakes. See Pro pricing.

60-second cheat sheet

The night-before-the-interview bullets. Scan these on the way to the call.

Sign in for 5 unlocks/day or go Pro for unlimited revision sheets. See Pro pricing.

References

Primary sources. Skim if you want the original framing.