AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

What Does Flow-Matching Bring to TD-Learning?

Bhavya Kumar Agrawalla, Michal Nauman, Aviral Kumar

ICML 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Flow matching generative modeling · 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
use distributional backup instead of expected-value backup (ablation)floq · OGBench hmmaze-largesuccess rate: 52 -> 30Hurt
use distributional backup instead of expected-value backup (ablation)floq · OGBench antmaze-giantsuccess rate: 86 -> 74Hurt
use distributional backup instead of expected-value backup (ablation)floq · OGBench cube-double and hmmaze-mediumsuccess rate: 72 -> 72; 94 -> 94No clear effect
stale velocity field for first 25% of integration (ablation)floq · OGBench antsoccer-arenasuccess rate: 45 -> 50Improved
stale velocity field for first 25% of integration (ablation)floq · OGBench cube-doublesuccess rate: 72 -> 84Improved
stale velocity field for first 25% of integration (ablation)floq · OGBench hmmaze-mediumsuccess rate: 98 -> 43Hurt
stale velocity field for all integration steps (100%) (ablation)floq · OGBench antsoccer-arenasuccess rate: 45 -> 35Hurt
freeze early layers of monolithic critic (ablation)FQL monolithic critic · OGBenchsuccess rateHurt
add i.i.d. noise to TD velocity targets, flow vs monolithic critic (ablation)floq · OGBenchsuccess rateImproved
freeze early features during TD learning (ablation)floq · OGBenchsuccess rateNo clear effect
train at a single integration step only (zero only) (ablation)floq · OGBenchstabilityHurt
supervise TD target directly instead of velocity at each step (ablation)floq · OGBenchsuccess rateHurt

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