Derives μP scalings for GQA via promoted spectral-norm definition of feature learning and a modified norm preserving scaling laws for non-full-rank matrices, with experiments showing learning-rate transfer.
International Conference on Machine Learning (ICML) , year=
12 Pith papers cite this work, alongside 1 external citations. Polarity classification is still indexing.
12
Pith papers citing it
1
external citations · external index
citation-role summary
background 2
citation-polarity summary
roles
background 2polarities
background 2representative citing papers
The authors derive a Maximally Scale-Stable Parameterization (MSSP) for MoE models that achieves robust learning-rate transfer and monotonic performance gains with scale across co-scaling regimes of width, experts, and sparsity.
ALU uses public data to suppress unlearning cost quadratically while characterizing distribution mismatch effects, enabling mass unlearning with maintained utility.
In a random feature model, optimal SGD learning-rate schedules are polynomial decay in the easy phase and warmup-stable-decay in the hard phase, outperforming constant or simple power-law schedules and transferring differently across training horizons.
A recurrent-depth architecture enables language models to improve reasoning performance by iterating computation in latent space, achieving gains equivalent to much larger models on benchmarks.
νGPT is a modified parameterization of normalized transformers that enables learning rate transfer across width, depth, and token horizon.
C-voting improves recurrent reasoning models by selecting among multiple latent trajectories the one with highest average top-1 probability, achieving 4.9% better Sudoku-hard accuracy than energy-based voting and outperforming HRM on Sudoku-extreme and Maze when paired with the new ItrSA++ model.
Parcae stabilizes looped LLMs via spectral norm constraints on injection parameters, enabling power-law scaling for training FLOPs and saturating exponential scaling at test time that improves quality over fixed-depth baselines under fixed parameter budgets.
Derives a novel two-point deterministic equivalence for random matrix resolvents to obtain unified asymptotics for SGD-trained linear regression, kernel regression, and random feature models.
GWT projects gradients into wavelet subspaces to compress optimizer states for memory-efficient LLM training while claiming performance parity with full-rank updates.
Splitting weight matrices into a fixed-norm direction and learnable per-row/column magnitudes improves LLM training over AdamW/Muon, removes weight decay and warmup, and transfers the optimal LR across width.
Review of neural scaling laws and their relation to constraints and inductive biases when applying machine learning to physics problems.
citing papers explorer
-
GQA-{\mu}P: The maximal parameterization update for grouped query attention
Derives μP scalings for GQA via promoted spectral-norm definition of feature learning and a modified norm preserving scaling laws for non-full-rank matrices, with experiments showing learning-rate transfer.
-
How to Scale Mixture-of-Experts: From muP to the Maximally Scale-Stable Parameterization
The authors derive a Maximally Scale-Stable Parameterization (MSSP) for MoE models that achieves robust learning-rate transfer and monotonic performance gains with scale across co-scaling regimes of width, experts, and sparsity.
-
Unlearning with Asymmetric Sources: Improved Unlearning-Utility Trade-off with Public Data
ALU uses public data to suppress unlearning cost quadratically while characterizing distribution mismatch effects, enabling mass unlearning with maintained utility.
-
Theory of Optimal Learning Rate Schedules and Scaling Laws for a Random Feature Model
In a random feature model, optimal SGD learning-rate schedules are polynomial decay in the easy phase and warmup-stable-decay in the hard phase, outperforming constant or simple power-law schedules and transferring differently across training horizons.
-
Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach
A recurrent-depth architecture enables language models to improve reasoning performance by iterating computation in latent space, achieving gains equivalent to much larger models on benchmarks.
-
Learning Rate Transfer in Normalized Transformers
νGPT is a modified parameterization of normalized transformers that enables learning rate transfer across width, depth, and token horizon.
-
C-voting: Confidence-Based Test-Time Voting without Explicit Energy Functions
C-voting improves recurrent reasoning models by selecting among multiple latent trajectories the one with highest average top-1 probability, achieving 4.9% better Sudoku-hard accuracy than energy-based voting and outperforming HRM on Sudoku-extreme and Maze when paired with the new ItrSA++ model.
-
Parcae: Scaling Laws For Stable Looped Language Models
Parcae stabilizes looped LLMs via spectral norm constraints on injection parameters, enabling power-law scaling for training FLOPs and saturating exponential scaling at test time that improves quality over fixed-depth baselines under fixed parameter budgets.
-
Two-Point Deterministic Equivalence for Stochastic Gradient Dynamics in Linear Models
Derives a novel two-point deterministic equivalence for random matrix resolvents to obtain unified asymptotics for SGD-trained linear regression, kernel regression, and random feature models.
-
GWT: Scalable Optimizer State Compression for Large Language Model Training
GWT projects gradients into wavelet subspaces to compress optimizer states for memory-efficient LLM training while claiming performance parity with full-rank updates.
-
Improving Neural Network Training by Decoupling the Magnitude and Direction of Weight Vectors
Splitting weight matrices into a fixed-norm direction and learnable per-row/column magnitudes improves LLM training over AdamW/Muon, removes weight decay and warmup, and transfers the optimal LR across width.
-
Statistical Properties of Training & Generalization
Review of neural scaling laws and their relation to constraints and inductive biases when applying machine learning to physics problems.