rl_sorting
Part of the RL module ecosystem; ships in DXPR CMS.
RL: A/B Test Views Content (rl_sorting) is a Views sort plugin that A/B and multivariate tests the order of any Drupal View using reinforcement learning. Every visitor click is human feedback (RLHF-style); the algorithm continuously shifts traffic to higher-performing items. No fixed test horizons. No manual winner picking. No third-party SaaS.
You need this if
- You want to A/B or multivariate test the order of content in a View (blog posts, products, news items, anything Views can list)
- You want to test dozens or hundreds of items simultaneously, not just A vs B
- You want continuous optimization, not fixed-horizon tests with manual setup
- You want fair exposure for new content without manually rotating it
Features
- Multivariate by default: 2 to thousands of items, no per-test setup
- Universal entity support: nodes, users, taxonomy terms, media, custom entities
- Views integration: a single sort plugin you add to any View display
- Real-time RLHF loop: clicks update the model on every page
- Cold-start handling: new items get exploration scores so they get fair exposure
- Time-based filtering: rolling windows for seasonal or campaign content
- Configurable caching: Views cache lifetime auto-aligned with the RL refresh rate
- Privacy-first tracking: anonymous statistics only, GDPR-compliant
How it works
The sort plugin scores every item in the View using a multi-armed bandit (Thompson Sampling) and routes traffic to higher-performing items. JavaScript on the rendered page records views (turns) and clicks (rewards), feeding the model in real time. The RL core handles the math; this module bridges it to Views.
Post-installation
- Install and enable the RL (Reinforcement Learning) module
- Configure permissions at
/admin/people/permissions - Edit any View display and add "RL: A/B Testing" as the sole sort criterion
- Configure cache refresh rate and time-window options
- Save; content optimization begins immediately
Prefer a turnkey demo site?
Spin up DXPR CMS, Drupal pre-configured with DXPR Builder, DXPR Theme, RL, and security best practices. DXPR CMS is a marketing CMS built for teams that need A/B testing out of the box.
Security & privacy
- GDPR-compliant: only anonymous interaction counts collected
- No personal data: no user identification, cookies, or per-individual tracking
- Local data storage: all data stays on your Drupal site, no external services
Migrating from ai_sorting
This module was renamed from ai_sorting in 2026. Existing installs migrate automatically when you enable rl_sorting; experiment history, Views config, and the legacy module are all handled in hook_install. See the 1.0.0-beta1 release notes for details.
Requires
- RL (Reinforcement Learning) module
- Drupal 10.3+ or 11
Related modules
- RL: A/B & Multivariate Testing (rl): required core engine. Also ships RL: A/B Test Page Titles and RL: A/B Test Menu Links as bundled submodules.
- Views: required Drupal core. rl_sorting registers itself as a Views sort plugin.
- DXPR CMS: ships rl_sorting pre-configured alongside DXPR Builder and DXPR Theme.