Tutorial
Rank and fuse multiple model outputs
Generate several candidate answers, rank them, and fuse the best parts into one result. Useful when quality matters more than speed.
Before you start
- An LLMWeave account
- Finished the consensus draft tutorial
Pick the rank-and-fuse template
Start a New Weave with the rank-and-fuse template. It generates multiple candidates, scores them, and fuses the top results.
Tune the candidate set
Decide how many candidates and which models generate them. A wider set costs more but gives the ranking step more to work with.
Run and inspect the ranking
The weave ranks candidates and fuses the strongest into one answer. The intermediate work is preserved so you can see why the final result was chosen.
Keep going
Build a consensus draft that merges multiple models
Instead of picking one model’s answer, merge several into a single stronger draft. This is the pattern most people want when they say "use multiple models".
Intermediate · 4 minGround a weave in current sources with web search
Turn on web search so models answer from current sources rather than training memory. Essential for anything recent or fact-sensitive.