Static HTML snapshot of the model record for crawlers and no-JS readers.
| Attribute | Value |
|---|---|
| Model | Value Prediction Network (VPN) |
| Lab / Organization | DeepMind |
| Category | Model-Based RL |
| Subtype | Value-Targeted Dynamics Model |
| World Model Type | Value-focused abstract dynamics |
| Primary Domain | Planning / Games |
| Architecture | Abstract state transition model with value and reward prediction heads |
| Modality | Visual / Abstract states |
| Training Method | End-to-end training with value and reward prediction targets |
| Status | foundational |
| Year | 2017 |
| Performance Index | 43/100 (medium confidence, v1.1) |
Main editorial body preserved directly in static HTML.
The Value Prediction Network learns abstract state dynamics optimized for predicting future rewards and values rather than reconstructing raw observations. It performs multi-step planning in abstract space, rolling forward predictions of value through learned transitions. This approach avoids the pitfalls of pixel-level prediction while retaining the benefits of model-based planning. VPN showed that models optimized directly for value prediction can outperform both model-free methods and models that predict observations.
Value Prediction Network (VPN) is a value-focused abstract dynamics developed by University of Michigan / Google Brain in 2017 for planning / games.
Short extractable facts for answer engines and no-JS readers.
| Signal | Value |
|---|---|
| Definition | Value Prediction Network (VPN) is a value-focused abstract dynamics developed by University of Michigan / Google Brain in 2017 for planning / games. |
| Short Description | A neural network that learns to plan by predicting future values and rewards through abstract state transitions, without reconstructing observations. |
| Benchmark Rows | 1 |
| FAQ Entries | 1 |
| Related Models | 2 |
| Related Guides | 0 |
| Related Research Topics | 1 |
| Last Updated | 2026-01-15 |
Key capabilities associated with this model.
Representative applications attached to this model record.
Balanced assessment surfaced in static HTML.
Published benchmark evidence attached to this model record.
| Benchmark | Metric | Result | Source |
|---|---|---|---|
| Atari | Mean HNS 1.08 x baseline | Improved over model-free baselines | Source |
Primary references preserved in static HTML for citation extraction.
| Reference | Link |
|---|---|
| Oh et al., 2017. Value Prediction Network. NeurIPS 2017. | Open source |
Nearby models linked from the current editorial record.
| Model | Category | World Model Type | Index v1.1 |
|---|---|---|---|
| Predictron | Model-Based RL | Abstract internal dynamics model | 43/100 |
| MuZero | Model-Based RL | Abstract learned dynamics + MCTS | 78/100 |
Side-by-side comparisons already connected to this model.
| Comparison | Matchup | Summary |
|---|---|---|
| Predictron vs MuZero | Predictron vs MuZero | Both learn abstract dynamics models for planning without requiring environment reconstruction, but Predictron was an early prototype while MuZero became the definitive realization of value-equivalent model learning. |
Connected research areas surfaced directly in static HTML.
| Topic | Summary |
|---|---|
| Model-Based Reinforcement Learning | What model-based reinforcement learning is, how world models enable imagination-based planning, and why Dreamer, MuZero, PlaNet, and TD-MPC2 matter. |
FAQ answers rendered directly into static HTML for extractable responses.
Both predict values in abstract space, but VPN explicitly learns state transitions and reward predictions along imagined trajectories, while the Predictron uses a single network with λ-weighted returns across depths.
Short extractable summary preserved directly in static HTML.
Editorial provenance and refresh policy preserved directly in static HTML.
Published by world-models.io editorial board.
Lead editor Tyler D. - Technical editor, methodology and benchmark analysis.
This model page synthesizes primary papers, official model pages, benchmark evidence, and related world-models.io context into a reference resource.
Each editorial page is assembled from primary sources, normalized into extractable summaries, checked for factual drift, and reviewed before publication or major refreshes. Last reviewed: 2026-01-15.
Pages are refreshed when a new paper, benchmark, release, architecture update, or stronger primary source materially changes the answer a reader or AI system should retrieve.
Each page links back to relevant primary sources and keeps a stable canonical URL so readers can verify claims, trace context, and reference the most up-to-date version. See the editorial policy.
Primary model and lab sources embedded in static HTML.