Loading patterns…
Process Reward Models & Verifier-Guided Search(PRM)
Uses a reward model that scores each intermediate reasoning step (process supervision) rather than only the final answer (outcome supervision). The per-step scores rank and prune candidate solutions during best-of-N sampling or tree search, concentrating test-time compute on the most promising branches and on harder problems.
In 30 seconds
- What
- Scores each reasoning step with a learned model, then prunes low-scoring solution branches early during search or sampling.
- When to use
- Multi-step problems where intermediate correctness matters, test-time compute is flexible, and you can afford to train a step-level verifier.
- Watch out
- Process reward model quality directly limits pruning accuracy; poor step scores waste compute or cut off correct paths prematurely.
Ask the AI expert about this pattern
Opens the assistant with your question prefilled. You review it before sending.
Process Reward Models & Verifier-Guided Search: Overview
Uses a reward model that scores each intermediate reasoning step (process supervision) rather than only the final answer (outcome supervision). The per-step scores rank and prune candidate solutions during best-of-N sampling or tree search, concentrating test-time compute on the most promising branches and on harder problems.
- Per-step correctness scoring (process supervision)
- Ranks and prunes candidates in best-of-N or tree search
- Early pruning of low-scoring branches
- Compute-optimal allocation by problem difficulty
- More sample-efficient than outcome-only reward models
- Verifier-guided selection of the final answer
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
References
The papers, specifications, and repositories this pattern is based on.
From the engineer behind this catalog
Get your agent architecture reviewed
This page documents one pattern. Your system runs dozens, and most failures live in how they fit together. Have the whole design reviewed against the 288 patterns in this catalog: architecture, reliability, evaluation and cost, every finding mapped to the pattern that fixes it.
€750 instead of €1,500, one week, written report and walkthrough call, until 30 September