BubbleSpec exploits long-tail bubbles in synchronous RL by using faster ranks' idle time to pre-generate rollout drafts for speculative decoding, reducing steps by 50% and raising throughput up to 1.8x while preserving exact synchrony.
Defeating the training-inference mismatch via fp16
10 Pith papers cite this work. Polarity classification is still indexing.
10
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
years
2026 10roles
background 2polarities
background 2representative citing papers
QaRL aligns quantized rollouts with training in LLM RL and uses TBPO with dual clipping to stabilize optimization, delivering +5.5 improvement over standard quantized-rollout baselines on Qwen3-30B math problems while retaining speed benefits.
PrecisionDiff is a differential testing framework that uncovers widespread precision-induced behavioral disagreements in aligned LLMs, including safety-critical jailbreak divergences across precision formats.
GDSD reduces RL for dLLMs to likelihood-free self-distillation via a normalization-free logit-matching objective, outperforming ELBO methods with more stable training on LLaDA-8B and Dream-7B.
NFPO augments the PPO surrogate with N-step forward traces to bridge local approximations and exact policy gradients, delivering tighter policy-improvement bounds and improved results on reasoning benchmarks.
Mu-GRPO enables substantially more off-policy GRPO training for LLMs via relaxed clipping and negative-advantage veto in large staged batches, matching standard GRPO performance at ~2x training speed.
A new RL objective adapts trust-region and off-policy handling automatically via normalized effective sample size of batch policy ratios, matching tuned baselines without new hyperparameters.
Rollout cards preserve complete agent rollout records and declare the reporting rules behind scores, enabling reproducible evaluation where changing only the rule can alter success rates by over 20 percentage points.
FP8-RL delivers up to 44% faster rollouts in LLM RL by using blockwise FP8 quantization, KV-cache recalibration, and importance-sampling corrections while keeping learning behavior close to BF16 baselines.
citing papers explorer
-
BubbleSpec: Turning Long-Tail Bubbles into Speculative Rollout Drafts for Synchronous Reinforcement Learning
BubbleSpec exploits long-tail bubbles in synchronous RL by using faster ranks' idle time to pre-generate rollout drafts for speculative decoding, reducing steps by 50% and raising throughput up to 1.8x while preserving exact synchrony.
-
QaRL: Rollout-Aligned Quantization-Aware RL for Fast and Stable Training under Training--Inference Mismatch
QaRL aligns quantized rollouts with training in LLM RL and uses TBPO with dual clipping to stabilize optimization, delivering +5.5 improvement over standard quantized-rollout baselines on Qwen3-30B math problems while retaining speed benefits.
-
Hidden Reliability Risks in Large Language Models: Systematic Identification of Precision-Induced Output Disagreements
PrecisionDiff is a differential testing framework that uncovers widespread precision-induced behavioral disagreements in aligned LLMs, including safety-critical jailbreak divergences across precision formats.
-
GDSD: Reinforcement Learning as Guided Denoiser Self-Distillation for Diffusion Language Models
GDSD reduces RL for dLLMs to likelihood-free self-distillation via a normalization-free logit-matching objective, outperforming ELBO methods with more stable training on LLaDA-8B and Dream-7B.
-
Multi-Step Likelihood-Ratio Correction for Reinforcement Learning with Verifiable Rewards
NFPO augments the PPO surrogate with N-step forward traces to bridge local approximations and exact policy gradients, delivering tighter policy-improvement bounds and improved results on reasoning benchmarks.
-
How Off-Policy Can GRPO Be? Mu-GRPO for Efficient LLM Reinforcement Learning
Mu-GRPO enables substantially more off-policy GRPO training for LLMs via relaxed clipping and negative-advantage veto in large staged batches, matching standard GRPO performance at ~2x training speed.
-
Trust the Batch, On- or Off-Policy: Adaptive Policy Optimization for RL Post-Training
A new RL objective adapts trust-region and off-policy handling automatically via normalized effective sample size of batch policy ratios, matching tuned baselines without new hyperparameters.
-
Rollout Cards: A Reproducibility Standard for Agent Research
Rollout cards preserve complete agent rollout records and declare the reporting rules behind scores, enabling reproducible evaluation where changing only the rule can alter success rates by over 20 percentage points.
-
FP8-RL: A Practical and Stable Low-Precision Stack for LLM Reinforcement Learning
FP8-RL delivers up to 44% faster rollouts in LLM RL by using blockwise FP8 quantization, KV-cache recalibration, and importance-sampling corrections while keeping learning behavior close to BF16 baselines.
- Cost-Aware Learning