AI Paper Atlas

Research map / Reinforcement learning / Offline RL

Flow-Based Single-Step Completion for Efficient and Expressive Policy Learning

Prajwal Koirala, Cody Fleming

ICLR 2026 · Reinforcement learning › Offline RL · approach: Offline RL sample complexity theory · 11 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
add single-step completion policy (SSCQL) in place of diffusion actor DQLDDPG+BC offline RL · D4RL MuJoCo averagenormalized score: 87.9 -> 87.9No clear effect
replace DQL's 5-step denoising actor with one-step SSCQLoffline RL actor · training and inference timetraining time: ~8 hours -> ~16 minsImproved
use one-step SSCQL instead of 5-step DQL denoisingoffline RL actor · inference latencyinference time (ms): 1.27 -> 0.27Improved
use SSCQL instead of DQL/CAC in fully online RLonline RL · Online RL averagenormalized score: 67.7 -> 90.53Improved
use SSCQL instead of DQL in offline-to-online finetuning on Hopper-MEoffline-to-online finetuning · Hopper-MEnormalized score: 71.7 -> 111.27Improved
increase SSCQL offline/online steps from 100K+100K to 250K+250K (ablation)SSCQL finetuning · Halfcheetah-MEnormalized score: 97.32 -> 110.68Improved
add single-step completion policy in place of Diffusion Policy for behavior cloningBC · RoboMimic/Push-T averagesuccess rate: 0.88 -> 0.87No clear effect
SSCP vs Diffusion Policy on ToolhangBC · Toolhangsuccess rate: 0.87 -> 0.64Hurt
add shared-network hierarchy distillation (GC-SSCP) in place of hierarchical HIQLgoal-conditioned offline RL · OGBench averagesuccess rate: 50.94 -> 54.44Improved
GC-SSCP vs HIQL on antmaze-large-stitchgoal-conditioned offline RL · antmaze-large-stitch-v0success rate: 67 -> 40Hurt
add completion loss (alpha_2 > 0) to flow matching loss (ablation)SSCQL actor · D4RL MuJoCo (Hopper, Walker2d medium)normalized scoreImproved
add completion loss on HalfCheetah-Medium (ablation)SSCQL actor · HalfCheetah-mediumnormalized scoreNo clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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