AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

Contrastive Active Inference

Pietro Mazzaglia, Tim Verbelen, Bart Dhoedt

NeurIPS 2021 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Intrinsically motivated exploration · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace likelihood (reconstruction) representation model with contrastive oneActive inference agent · representation model computeMMACs: 212.2 -> 15.4Improved
replace likelihood representation model with contrastive oneActive inference agent · representation model size# Params: 4485.7k -> 1266.7kImproved
use contrastive instead of reconstruction-based agentDreamer / AIF · wall-clock time vs Dreamerrelative time: 1.00 -> 0.84Improved
use observation reconstruction for AIF instead of contrastiveActive inference agent · wall-clock time vs Dreamerrelative time: 0.84 -> 3.24Hurt
use contrastive AIF instead of likelihood AIFActive inference agent · MiniGrid Empty 8x8reward consistencyImproved
use contrastive AIF instead of likelihood AIFActive inference agent · MiniGrid Empty 6x6mean reward: below 0.4 -> comparable to MBRLImproved
use contrastive AIF instead of DreamerActive inference agent · Reacher Easy and Hardconvergence speedImproved
use likelihood AIF instead of contrastive AIFActive inference agent · Reacher Easy/HardrewardHurt
use Dreamer (reconstruction) instead of contrastive modelsMBRL · Distracting Reacher EasyrewardHurt
use contrastive Dreamer on distracting backgroundsDreamer · Distracting Reacher EasyrewardImproved
use Contrastive-AIF with unchanged goal image under background changeActive inference agent · Distracting Reacher EasyrewardImproved
use contrastive utility instead of likelihood utilityActive inference agent · Reacher Hard utility valuesutility discrimination: 0.98 vs 1.00Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy