AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

FAIL: Flow Matching Adversarial Imitation Learning for Image Generation

Yeyao Ma, Chen Li, Xiaosong Zhang, Han Hu, Weidi Xie

ICML 2026 · Reinforcement learning › Imitation learning · approach: Adversarial imitation learning · 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 FAIL-PG instead of SFT on expert dataFLUX.1-dev · UniGen-Bench (512px)score: 58.22 -> 64.75Improved
train RLHF-PD for 400 instead of 50 steps (ablation)RLHF-PD · UniGen-Benchscore: 59.76 -> 57.97Hurt
replace Online DPO with FAIL-PG (learned discriminator)FLUX.1-dev · UniGen-Benchscore: 62.83 -> 64.75Improved
extend FAIL-PG training beyond 450 steps (ablation)FAIL-PG · UnifiedReward over trainingrewardHurt
extend FAIL-PD training to 2,000 steps (ablation)FAIL-PD · UnifiedReward over trainingreward: 3.4759 at step 2,000Improved
use DINOv3 visual foundation model discriminator instead of Qwen3-VL (ablation)FAIL-PD · UniGen-Benchscore: 62.17 -> 60.65Hurt
use FLUX.1 backbone discriminator instead of Qwen3-VL (ablation)FAIL-PD · HPSv3score: 10.82 -> 12.35Improved
vary discriminator type (DINOv3, FLUX.1, Qwen3-VL) (ablation)FAIL-PG · UniGen-Benchscore: 64.23, 63.33, 64.75No clear effect
add FAIL-PD loss to Reward Gradient optimizationReward Gradient (HPSv3) · UniGen-Bench / UnifiedRewardscore: 58.50 -> 60.88 (UniGen); 3.4254 -> 3.4832 (URScore)Improved
add FAIL-PG score to FPOFPO (HPSv3 reward) · UniGen-Benchscore: 61.27 -> 64.27Improved
apply FAIL-PG to Xomni instead of SFTXomni (discrete AR) · UniGen-Benchscore: 41.36 -> 54.24Improved
apply FAIL-PG instead of SFT to Wan2.1-1.3BWan2.1-1.3B · VBench Semantic Scorescore: 46.27 -> 51.75Improved

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