Crimson AI NewsA CrimsonLingua Network service
EN ع
← Back to news
Research paper Hugging Face

WarpSAC: Regime-Aware Off-Policy RL Boosts Scalability and Sim-to-Real Transfer

AI By Crimson AI Hugging Face Papers 27 August 2026 · 00:00 1 views
Share: X Telegram

Hugging Face researchers introduce WarpSAC, a family of off-policy RL algorithms that adapt stabilizers to data availability, improving efficiency across CPU and GPU-parallel training and achieving significant gains in sim-to-real deployment.

WarpSAC: Regime-Aware Off-Policy RL Boosts Scalability and Sim-to-Real Transfer

Key points

Off-policy reinforcement learning (RL) has long relied on stabilizers designed for data-limited replay buffers. However, with the rise of massively parallel simulation, the data regime has shifted, challenging these stabilizers' effectiveness. Researchers at Hugging Face have conducted controlled experiments across eight benchmark families to investigate this issue.

Their findings reveal that stabilizers are data-regime-dependent: parameter normalization helps when replay coverage is narrow but restricts value fitting when data are abundant, while clipped double-Q can be relaxed in high-throughput manipulation. Age-biased replay weighting improves learning efficiency across regimes, especially with limited network capacity.

Based on these insights, the team proposes WarpSAC, a regime-aware family of off-policy RL algorithms. WarpSAC introduces Sample Weight Decay for efficient exploitation and offers two variants: WarpSAC-L (Norm ON, clipped double-Q) for data-limited CPU-scale training, and WarpSAC-A (Norm OFF, single-Q) for data-abundant GPU-parallel training.

In evaluations, WarpSAC improves normalized score-step AUC over FlashSAC by 4.5% across nine CPU-scale environments and 23.1% across fourteen GPU-parallel environments. It also increases the UnitreeG1TransportBox-v1 success rate from 19.8% to 96.4%, improves mean normalized wall-time AUC on MuJoCo Playground by 19.1%, and achieves 36.4% faster sim-to-real deployment on Unitree G1 compared to FlashSAC.

The project page provides an overview, demonstrations, benchmark results, and sim-to-real evaluations, while the GitHub repository contains source code, installation instructions, training scripts, environment integrations, and experiment configurations.

MetricWarpSAC Improvement
Normalized score-step AUC (CPU-scale, 9 envs)+4.5% vs FlashSAC
Normalized score-step AUC (GPU-parallel, 14 envs)+23.1% vs FlashSAC
UnitreeG1TransportBox-v1 success rate19.8% → 96.4%
Mean normalized wall-time AUC (MuJoCo Playground)+19.1%
Sim-to-real deployment speed (Unitree G1)36.4% faster than FlashSAC
Source
Hugging Face · Hugging Face Papers
Related news
Research paper
Hugging Face 29 Aug 2026

Hugging Face Audit: 110 of 124 AI Evaluations Fail to Support Their Claims

A new commit-bound census of 124 Inspect Evals units reveals that 110 stop before deterministic inference due to missing historica...

4
Research paper
Hugging Face 29 Aug 2026

Aphanta: New Framework Diagnoses When Image Editing Boosts Multimodal Reasoning

Hugging Face researchers introduce Aphanta, a diagnostic framework that evaluates when image-editing intermediates improve multimo...

5
Research paper
Hugging Face 29 Aug 2026

EditaLive! Enables Real-Time Character Video Editing for Live Streaming

Hugging Face researchers introduce EditaLive, a framework for real-time human-centric video editing in live streams, achieving sta...

4