6arXiv cs.LG (Machine Learning)·
Parallel Decoding Distillation (PDD) accelerates video and image diffusion inference with improved diversity
Researchers introduce Parallel Decoding Distillation (PDD), a trajectory-based distillation method that accelerates diffusion and flow matching models by predicting multiple denoising steps per network evaluation, achieving state-of-the-art performance at 4-8 NFE. Unlike prior methods relying on variational score distillation and adversarial losses, PDD avoids mode collapse and improves generated video diversity. The method is validated on LTX-2.3 Text-to-Video/Audio, Wan 14B Text-to-Video, and Qwen-Image Text-to-Image models.
Related guides (2)
Related events (8)
On-Policy Delta Distillation improves reasoning transfer via teacher-base model difference signal
Researchers from NAVER AI introduce On-Policy Delta Distillation (OPD²), a new post-training method that replaces direct imitation of a teacher model's output distribution with a 'delta signal' — the difference between the teacher and its pre-instruction-tuning base model. This delta signal isolates the reasoning capability changes induced by instruction tuning, providing a more targeted supervision signal for student models. Experiments across math, science, and code-reasoning benchmarks show OPD² consistently outperforms conventional on-policy distillation with shorter post-training periods.
ADAS: Attention-Discounted Adaptive Sampler improves parallel decoding for masked diffusion language models
Researchers propose ADAS, a training-free reranking rule for masked diffusion language model decoding that addresses token interaction failures in parallel token commitment. The method greedily penalizes candidates that attend strongly to already-selected uncertain positions, using attention weights as soft marginal penalties rather than hard constraints. Evaluated on LLaDA-8B-Base and Dream-7B-Base across GSM8K, MATH500, HumanEval, and MBPP, ADAS improves low-NFE performance by 9–10 percentage points on average when plugged into existing samplers with only 3.1% runtime overhead.
TunerDiT: Training-free Progressive Steering of Diffusion Transformers for Multi-Event Video Generation
TunerDiT is a training-free method for steering video diffusion transformers (DiTs) to generate long-horizon videos containing multiple sequential events. The approach identifies intrinsic turning points in the DiT denoising trajectory where text conditioning shifts from global layout to fine-grained detail, then applies two steering mechanisms: Event-Partitioned Masking and Cross-Event Prompt Fusion. The authors also introduce Meve, a benchmark prompt suite for multi-event video generation, and report state-of-the-art results across 8 metrics with improved text alignment scaling with event count.
Finetune Stable Diffusion Models with DDPO via TRL
Hugging Face's TRL library adds support for DDPO (Denoising Diffusion Policy Optimization), enabling reinforcement learning-based finetuning of Stable Diffusion models. This extends TRL's RLHF tooling beyond language models to image generation, allowing reward-driven optimization of diffusion models. The post demonstrates practical usage of the new DDPO trainer within the TRL ecosystem.
Relay-OPD addresses prefix failure in on-policy distillation via teacher-student trajectory handoff
Researchers introduce Relay On-Policy Distillation (Relay-OPD), a training method that addresses 'prefix failure' in on-policy knowledge distillation, where student models compound early reasoning errors throughout a trajectory. The approach detects divergence points where teacher and student continuations asymmetrically diverge, then briefly hands generation to the teacher to produce a corrective 'relay leg' before the student resumes. Evaluated on eight mathematical reasoning benchmarks using Qwen3-4B-Instruct-2507 as teacher and Qwen3-0.6B/1.7B as students, Relay-OPD outperforms standard OPD by +5.73% and the strongest baseline FastOPD by +1.49% on average for the 1.7B model, while also reducing training trajectory length by over 50%.
AdaFlash: Adaptive speculative decoding with on-policy distilled diffusion drafters
AdaFlash is a new framework for accelerating LLM inference via speculative decoding that addresses variance problems in diffusion-based draft models. It combines an on-policy distillation algorithm using reverse-KL divergence with an adaptive length head that dynamically adjusts candidate sequence length. Experiments show up to ~66% higher throughput than prior state-of-the-art methods, with especially large gains in high-concurrency deployment scenarios.
Analysis of on-policy distillation reveals sparse, geometrically structured parameter updates
A new arXiv paper analyzes on-policy distillation (OPD) — a post-training method combining on-policy student trajectories with dense teacher supervision — across language and vision-language model pairs. The authors find that OPD updates are coordinate-sparse and distributed across layers (FFN-heavy), and that training only the discovered sparse subnetwork recovers near-full performance. Geometrically, updates are numerically full-rank but spectrally concentrated, falling disproportionately on near-zero weight coordinates, suggesting OPD retains distinct geometric signatures rather than behaving like ordinary dense parameter rewriting.
DiT-Reward converts text-to-image diffusion transformers into reward models, outperforming HPSv3
DiT-Reward is a new reward modeling approach that repurposes pretrained text-to-image Diffusion Transformers (DiTs) by processing near-clean image latents and aggregating text-conditioned representations across transformer layers. Under matched training data, it outperforms HPSv3 on all four evaluated preference benchmarks, reaching 85.6% on HPDv2 and 77.6% on HPDv3. When used to optimize Stable Diffusion 3.5 Large via Flow-GRPO, it shows clear gains in realism and achieves a 1.65x inference speedup over HPSv3. The work demonstrates that generative DiT representations transfer meaningfully to reward modeling and policy optimization.