QUADS: Stabilizing NVFP4 Reinforcement Learning for MoE via QUantization-error Alignment across Dual Sides

• URL: http://arxiv.org/abs/2607.15810v1
• Date: Fri, 17 Jul 2026 10:21:27 GMT
• Title: QUADS: Stabilizing NVFP4 Reinforcement Learning for MoE via QUantization-error Alignment across Dual Sides
• Abstract summary: Rollout generation is a major bottleneck in Reinforcement Learning for Mixture-of-Experts (MoE) Large Language Models.<n>We propose QUantization-error Alignment across Dual Sides (QUADS) to stabilize NVFP4 RL for MoE.<n>QUADS achieves BF16-level accuracy, improves average pass@1 by 21.49 points over naive NVFP4 RL, and delivers 16% higher rollout throughput than FP8.
• Score: 26.939321070753607
• License: http://creativecommons.org/publicdomain/zero/1.0/
• Abstract: Rollout generation is a major bottleneck in Reinforcement Learning (RL) for Mixture-of-Experts (MoE) Large Language Models, motivating low-precision rollout acceleration such as FP8. As an emerging low-precision format, NVFP4 combines fine-grained scaling for accuracy preservation with native W4A4 FP4 GEMMs for higher throughput than FP8. However, we find that directly applying NVFP4 to MoE RL rollout is impractical. NVFP4 rollout with BF16 training collapses after roughly 150 steps, accompanied by rapidly growing rollout-trainer log-probability gaps. Through training-inference error analysis and controlled ablations, we identify activation error, rather than weight error, as the dominant source of FP4 RL instability: weights can be synchronized and aligned by a shared quantization-dequantization path, whereas activations are recomputed online and error is amplified by the coarse E2M1 grid. Therefore, to stabilize NVFP4 RL for MoE, we propose QUantization-error Alignment across Dual Sides (QUADS). On the trainer side, we introduce Asymmetric Quantization-Aware Training fake-quantizing weights while keeping activations unquantized for better alignment. On the rollout side, Residual Activation Compensation corrects high-error activation channels while preserving native W4A4 GEMMs. In our MoE RL experiments on several benchmarks, QUADS achieves BF16-level accuracy, improves average pass@1 by 21.49 points over naive NVFP4 RL, and delivers ~16% higher rollout throughput than FP8.

Related papers

• HiFloat4 Format for End-To-End Reinforcement Learning Post-Training of Large Language Models [9.149747042283197]
  We present the first end-to-end FP4 RL post-training, in which both the rollout and training policies operate at 4-bit precision.<n>A lightweight correction, Rollout-ResQ, recovers most of the precision lost to outlier-driven underflow without inflating the rollout's compute footprint.
  arXiv  Detail & Related papers  (2026-07-29T06:28:26Z)
• Characterizing the Impact of NVFP4 Quantization for Low-Power Edge AI Deployment [9.460818703756205]
  Energy-efficient neural-network inference at the edge requires reducing arithmetic cost, memory traffic, energy, and storage overhead while maintaining acceptable accuracy.<n>This paper presents an ablation-focused study of NVFP4 quantization for edge-efficient neural networks.
  arXiv  Detail & Related papers  (2026-06-03T03:09:59Z)
• TORQ: Two-Level Orthogonal Rotation for MXFP4 Quantization [13.984796236174331]
  We propose TORQ (Two-level Orthogonal Rotation for MXFP4 Quantization) as a training-free Post-Training Quantization framework.<n>We show that TORQ significantly improves the accuracy of MXFP4 activation quantization compared to existing methods.
  arXiv  Detail & Related papers  (2026-05-19T09:05:47Z)
• FP4 Explore, BF16 Train: Diffusion Reinforcement Learning via Efficient Rollout Scaling [38.64059734487925]
  Reinforcement-Learning-based post-training has emerged as a promising paradigm for aligning text-to-image diffusion models with human preferences.<n> scaling rollouts on large-scale foundational diffusion models (e.g., FLUX.1-12B) imposes a heavy computational burden.<n>We propose Sol-RL, a novel FP4-empowered Two-stage Reinforcement Learning framework.
  arXiv  Detail & Related papers  (2026-04-08T10:14:47Z)
• Dissecting Outlier Dynamics in LLM NVFP4 Pretraining [46.10969678564592]
  This study conducts a longitudinal analysis of outlier dynamics across architecture during NVFP4 pretraining.<n>We find that, compared with Softmax Attention (SA), Linear Attention (LA) reduces per-tensor heavy tails but still exhibits persistent block-level spikes under block quantization.<n>We then develop CHON, an NVFP4 training recipe integrating with post-QK operation protection.
  arXiv  Detail & Related papers  (2026-02-02T12:50:27Z)
• Jet-RL: Enabling On-Policy FP8 Reinforcement Learning with Unified Training and Rollout Precision Flow [48.48936574810267]
  We present the first comprehensive study of FP8 RL training.<n>We propose Jet-RL, an FP8 RL training framework that enables robust and stable RL optimization.
  arXiv  Detail & Related papers  (2026-01-20T18:54:31Z)
• Pretraining Large Language Models with NVFP4 [53.235038214986865]
  We introduce a novel approach for stable and accurate training of large language models (LLMs) using the NVFP4 format.<n>Our method integrates two-dimensional quantization scheme for consistent representations across both the forward and backward passes.<n>Our results show that the model trained with our NVFP4-based pretraining technique achieves training loss and downstream task accuracies comparable to an FP8 baseline.
  arXiv  Detail & Related papers  (2025-09-29T17:53:17Z)
• Bridging the Gap Between Promise and Performance for Microscaling FP4 Quantization [77.67818998672516]
  We present the first comprehensive study of MXFP4 and NVFP4 for post-training quantization.<n>We introduce Micro-Rotated-GPTQ (MR-GPTQ), a variant of the classic GPTQ quantization algorithm.<n>We show that MR-GPTQ matches or outperforms state-of-the-art accuracy.
  arXiv  Detail & Related papers  (2025-09-27T09:22:21Z)
• Towards Fully FP8 GEMM LLM Training at Scale [77.97607456493257]
  Existing approaches often rely on suboptimal fine-grained FP8 kernels or fall back to higher-precision matrix multiplications.<n>We introduce a new class of LLM architectures that, for the first time, support FP8 computation for all GEMMs within transformer blocks during both forward and backward passes.<n>This enables unprecedented throughput gains, particularly at scale, while matching the downstream performance of standard BF16 training.
  arXiv  Detail & Related papers  (2025-05-26T21:04:14Z)
• Optimizing Large Language Model Training Using FP4 Quantization [73.55459961002371]
  Quantized training presents a promising solution by enabling low-bit arithmetic operations to reduce costs.<n>This work introduces the first FP4 training framework for large language models (LLMs)
  arXiv  Detail & Related papers  (2025-01-28T18:04:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.