FP4 Explore, BF16 Train: Diffusion Reinforcement Learning via Efficient Rollout Scaling research paper by NVIDIA, 2026

NVIDIA · Inference and efficiency · Apr 8, 2026 · 3 citations · 29 upvotes · unverified

A novel two-stage reinforcement learning framework called Sol-RL integrates FP4 quantization with diffusion model alignment to accelerate training while maintaining high-fidelity performance.

By Yitong Li, Junsong Chen, Shuchen Xue and 8 more · arXiv 2604.06916 · PDF

UnverifiedHugging Face's summary; not yet checked by hand.

More from NVIDIA

All 90

Multimodal and roboticsUnverifiedNov 5, 2026VFlash: Fast, Accurate Visual Planning via Flash Primitive DecodingA method called VFlash speeds up and improves visual motion planning for robots by predicting entire spatial shapes at once.

NVIDIA109 upvotesRead →Foundation modelsUnverifiedOct 27, 2026Nemotron-Labs-Diffusion-Uni: Elastic Self-Refining Unified Diffusion LLM with Cross-Paradigm On-Policy DistillationNLD-Uni is a unified discrete diffusion language model that combines visual understanding and generation via cross-paradigm on-policy distillation and progressive self-refining diffusion.

NVIDIA12 upvotesRead →ReasoningUnverifiedOct 27, 2026ThinkWorld: World Imagination Reinforces Chain-of-Dynamics ReasoningThinkWorld improves robotic generalization by representing actions as dynamics chains and using world imagination to learn from unlabeled human videos.

NVIDIA57 upvotesRead →Multimodal and roboticsUnverifiedOct 20, 2026Joints, Camera, Action! Co-Generating 3D Motion and Video for Arbitrary RigsThis model generates both video and 3D motion for any character rig, allowing the motion to be replayed accurately within a 3D scene.

NVIDIA26 upvotesRead →Inference and efficiencyUnverifiedOct 13, 2026EdgeWise: Bringing Long-Horizon Agents to the Edge via Selective Cloud DelegationThis paper introduces a framework that reduces inference costs by using a small local model for most tasks and delegating only critical steps to the cloud.

NVIDIA30 upvotesRead →Training and scalingUnverifiedOct 8, 2026When Do We Need On-Policy Distillation? Distilling on Offline Student Rollouts Is Often BetterThis paper shows that distilling from offline student rollouts is often more efficient and accurate than standard on-policy distillation for many model pairs.

NVIDIA28 upvotesRead →Multimodal and roboticsUnverifiedOct 7, 2026Long-WAM: Scaling the Context of World-Action ModelsLong-WAM improves robot control by using autoregressive pretraining to effectively leverage long visual histories for real-time action prediction.

NVIDIA2 citations114 upvotesRead →Inference and efficiencyUnverifiedOct 6, 2026NeMo-DCR: Bit-Exact Delta-Compressed Refit for Scalable Agentic RL at Trillion-Parameter ScaleNeMo-DCR enables efficient, bit-exact synchronization of trillion-parameter model checkpoints by transmitting only the sparse weight changes between clusters.

NVIDIA1 citations29 upvotesRead →

Topic

Inference and efficiencySep 12, 2023Efficient Memory Management for Large Language Model Serving with PagedAttentionvLLM manages the KV cache like pages of virtual memory, serving models with 2 to 4 times the throughput.

UC Berkeley9,965 citations72 upvotesRead →Inference and efficiencyMay 27, 2022FlashAttention: Fast and Memory-Efficient Exact Attention with IO-AwarenessFlashAttention computes exact attention with far fewer GPU memory reads and writes, making long sequences faster.

Stanford University5,830 citations16 upvotesRead →Inference and efficiencyUnverifiedJul 24, 2025Group Sequence Policy OptimizationGroup Sequence Policy Optimization (GSPO) is a reinforcement learning algorithm that improves training efficiency and performance of large language models by using sequence-level importance ratios and operations.

Alibaba (Qwen)814 citations323 upvotesRead →Inference and efficiencyUnverifiedFeb 16, 2025Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse AttentionNSA, a trainable sparse attention mechanism, enhances long-context modeling efficiency without sacrificing performance, achieving improvements in speed and accuracy over full attention models.

DeepSeek596 citations157 upvotesRead →Inference and efficiencyUnverifiedApr 7, 2025SmolVLM: Redefining small and efficient multimodal modelsSmolVLM, a series of compact multimodal models, achieves high performance with minimal GPU memory usage, making efficient deployment on mobile and edge devices possible.

Hugging Face330 citations212 upvotesRead →Inference and efficiencyUnverifiedOct 17, 2025Chronos-2: From Univariate to Universal ForecastingChronos-2, a pretrained model with a group attention mechanism, achieves state-of-the-art performance in zero-shot univariate, multivariate, and covariate-informed forecasting tasks.

Amazon276 citations28 upvotesRead →

About this paper

Venue: arXiv.org
Citations: 3, 0 influential · Semantic Scholar
Upvotes: 29 · Hugging Face
Lab: NVIDIA · on Companies · on Acquisitions · on Quarterly · on Paydays · on TechConf

Changes

CitationsNov 2, 2026Citations rose from 2 to 3.

CitationsOct 5, 2026Citations rose from 1 to 2.

CitationsSep 25, 2026A first count of 0 influential citations.

CitationsSep 25, 2026A first count of 1 citations.

New paperSep 25, 2026Found by the weekly Hugging Face scan; not yet checked by hand.

What it shows

More from NVIDIA

Other inference and efficiency papers

Changes