SGMD uses fake-score optimization toward the teacher with stop-gradient Fisher objective and NR/RC dual potentials to deliver ~3x training speedup and better motion dynamics in 4-step video diffusion models.
Flash-dmd: Towards high-fidelity few-step image generation with efficient distillation and joint reinforcement learning
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
years
2026 5verdicts
UNVERDICTED 5roles
background 2representative citing papers
RTDMD unifies KL minimization to a reward-tilted teacher into distribution matching plus reward terms, using AC-DMD in stage one and hybrid GRPO-style gradients plus SubGRPO in stage two to reach new SOTA on preference, aesthetic, and compositional metrics with 4-step generation on SD3, SD3.5, and F
A multi-teacher distillation framework that packs 50 effect LoRAs and fast sampling into a single adapter while aiming to avoid concept interference.
Super-Linear Advantage Shaping (SLAS) introduces a non-linear geometric policy update for RL post-training of text-to-image models that reshapes the local policy space via advantage-dependent Fisher-Rao weighting to reduce reward hacking and improve performance over GRPO baselines.
The paper introduces the Proxy Compression Hypothesis as a unifying framework explaining reward hacking in RLHF as an emergent result of compressing high-dimensional human objectives into proxy reward signals under optimization pressure.
citing papers explorer
-
SGMD: Score Gradient Matching Distillation for Few-Step Video Diffusion Distillation
SGMD uses fake-score optimization toward the teacher with stop-gradient Fisher objective and NR/RC dual potentials to deliver ~3x training speedup and better motion dynamics in 4-step video diffusion models.
-
Reinforcing Few-step Generators via Reward-Tilted Distribution Matching
RTDMD unifies KL minimization to a reward-tilted teacher into distribution matching plus reward terms, using AC-DMD in stage one and hybrid GRPO-style gradients plus SubGRPO in stage two to reach new SOTA on preference, aesthetic, and compositional metrics with 4-step generation on SD3, SD3.5, and F
-
CollectionLoRA: Collecting 50 Effects in 1 LoRA via Multi-Teacher On-Policy Distillation
A multi-teacher distillation framework that packs 50 effect LoRAs and fast sampling into a single adapter while aiming to avoid concept interference.
-
Power Reinforcement Post-Training of Text-to-Image Models with Super-Linear Advantage Shaping
Super-Linear Advantage Shaping (SLAS) introduces a non-linear geometric policy update for RL post-training of text-to-image models that reshapes the local policy space via advantage-dependent Fisher-Rao weighting to reduce reward hacking and improve performance over GRPO baselines.
-
Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges
The paper introduces the Proxy Compression Hypothesis as a unifying framework explaining reward hacking in RLHF as an emergent result of compressing high-dimensional human objectives into proxy reward signals under optimization pressure.