Pion modifies Muon's Newton-Schulz iterations into a controllable high-pass filter that anchors dominant singular values at 1 while suppressing noisy tails, outperforming Muon and AdamW in VLA and RLVR regimes.
Length-unbiased sequence policy optimization: Revealing and controlling response length variation in rlvr.arXiv preprint arXiv:2602.05261
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 4roles
background 1polarities
background 1representative citing papers
Replacing the importance sampling ratio with a stop-gradient self-anchored ratio for positive advantages yields unclipped, REINFORCE-equivalent gradients that improve exploration without training instability.
ICR creates a virtual shorter distribution from shortest correct on-policy responses to regularize RL post-training toward concise yet accurate reasoning, improving the accuracy-length Pareto frontier on math and knowledge benchmarks.
EqLen reframes length bias in sequence-level RL as a comparison-unit construction problem and builds equal-length training segments via dual-track generation, prefix inheritance, and segment masking.
citing papers explorer
-
Rethinking Muon Beyond Pretraining: Spectral Failures and High-Pass Remedies for VLA and RLVR
Pion modifies Muon's Newton-Schulz iterations into a controllable high-pass filter that anchors dominant singular values at 1 while suppressing noisy tails, outperforming Muon and AdamW in VLA and RLVR regimes.
-
UP: Unbounded Positive Asymmetric Optimization for Breaking the Exploration-Stability Dilemma
Replacing the importance sampling ratio with a stop-gradient self-anchored ratio for positive advantages yields unclipped, REINFORCE-equivalent gradients that improve exploration without training instability.
-
Implicit Compression Regularization: Concise Reasoning via Internal Shorter Distributions in RL Post-Training
ICR creates a virtual shorter distribution from shortest correct on-policy responses to regularize RL post-training toward concise yet accurate reasoning, improving the accuracy-length Pareto frontier on math and knowledge benchmarks.
-
Rethinking the Comparison Unit in Sequence-Level Reinforcement Learning: An Equal-Length Paired Training Framework from Loss Correction to Sample Construction
EqLen reframes length bias in sequence-level RL as a comparison-unit construction problem and builds equal-length training segments via dual-track generation, prefix inheritance, and segment masking.