Chain-of-thought prompting, by including intermediate reasoning steps in few-shot examples, elicits strong reasoning abilities in large language models on arithmetic, commonsense, and symbolic tasks.
super hub Canonical reference
Scaling Laws for Neural Language Models
Canonical reference. 84% of citing Pith papers cite this work as background.
851
Pith papers citing it
Background
84% of classified citations
abstract
We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence.
hub tools
citation-role summary
background 121
method 6
dataset 3
baseline 2
other 2
citation-polarity summary
claims ledger
- abstract We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are s
authors
co-cited works
representative citing papers
BBO-Pile is the first large-scale open dataset of real optimization trajectories used to train and scale foundation models that imitate black-box optimization methods.
Introduces the Synthetic Data Contamination Equilibrium and derives closed-form optimal provenance subsidies s* = KL(q||p)/(2 kappa) plus watermark strengths to mitigate model collapse, validated by OLS matching structural predictions on C4 data.
Cayley unitary adapters executed on real quantum hardware improve LLM perplexity by 1.4% on Llama 3.1 8B with 6000 parameters and recover 83% of compression-induced degradation on SmolLM2.
ε-coresets for attention exist of size O(√d e^{ρ+o(ρ)}/ε) for unit-norm keys/values and queries of norm ≤ρ, nearly matching the Ω(√d e^ρ/ε) lower bound.
RoundPipe achieves near-zero-bubble pipeline parallelism for LLM training on consumer GPUs by dynamically dispatching computation stages round-robin, yielding 1.48-2.16x speedups and enabling 235B model fine-tuning on 8x RTX 4090.
Transformer weight spectra exhibit transient compression waves that propagate layer-wise, persistent non-monotonic depth gradients in power-law exponents, and Q/K-V asymmetry, with the spectral exponent alpha predicting layer importance and enabling pruning gains of 1.1x-3.6x over Last-N baselines.
Grokking reflects escape from a metastable low-dimensional regime where transverse curvature accumulates before generalization, with subspace motion necessary but curvature boost insufficient.
The SDE benchmark shows LLMs lag on scientific discovery tasks relative to general science tests, with diminishing scaling returns and shared weaknesses across models.
Transformers perform kernel-based prediction for Hölder regression on manifolds and achieve intrinsic-dimension-dependent minimax rates with sufficient training tasks.
Introduces hybrid noise and novel coupling analysis to achieve the first convergent hidden-state DP bound for zeroth-order optimization.
LLaDA is a scalable diffusion-based language model that matches autoregressive LLMs like LLaMA3 8B on tasks and surpasses GPT-4o on reversal poem completion.
ErrorRadar is a new benchmark of 2,500 multimodal K-12 math problems for MLLM error step identification and categorization, where GPT-4o trails human experts by ~10%.
TTT layers treat the hidden state as a trainable model updated at test time, allowing linear-complexity sequence models to scale perplexity reduction with context length unlike Mamba.
KANs with learnable univariate spline activations on edges achieve better accuracy than MLPs with fewer parameters, faster scaling, and direct visualization for scientific discovery.
Tiny language models under 10M parameters trained on a synthetic children's story dataset generate fluent, consistent, multi-paragraph English text with near-perfect grammar and reasoning.
Pythia releases 16 identically trained LLMs with full checkpoints and data tools to study training dynamics, scaling, memorization, and bias in language models.
Language models can automatically generate high-quality evaluation datasets that reveal new cases of inverse scaling, sycophancy, and concerning goal-seeking behaviors, including some worsened by RLHF.
The Pile is a newly constructed 825 GiB dataset from 22 diverse sources that enables language models to achieve better performance on academic, professional, and cross-domain tasks than models trained on Common Crawl variants.
HERMES provides a reusable hierarchical labeling substrate for pre-training data that reveals granularity-specific effects in data mixing rules during model training.
Multi-agent LLMs generate and verify 14,073 deterministic reaction rules from 665,901 patents, enabling 97.7% classification of unseen reactions with finer resolution than fixed proprietary systems.
SemRF supplies fixed semantic anchors and pseudo-inverse tying to produce stable coordinates for residual dynamics, Voronoi traces, and minimum-action canonical paths that link to parameter efficiency under controlled interface error.
Cortex uses an Ontological Corpus Graph to structure web-scale corpora, creating a refined 24.14B-token corpus and a new benchmark validated on eight LLMs.
Token loss trajectories follow localized sigmoids whose learning-time spectrum quantitatively reconstructs scaling-law derivatives on T, D, and M axes and enables faster training via distribution reshaping.
citing papers explorer
-
Optimal Decay Spectra for Linear Recurrences
PoST reparameterizes decay spectra in linear recurrences with geometric log-spacing and position-adaptive scaling to achieve O(exp(-cN/log t)) decay, improving zero-shot language modeling and long-context retrieval across Mamba-2, RWKV-7 and similar models.
-
PoM: A Linear-Time Replacement for Attention with the Polynomial Mixer
PoM is a new linear-complexity token mixer using learned polynomials that matches attention performance in transformers while enabling efficient long-sequence processing.
-
CAKE: Cloud Architecture Knowledge Evaluation of Large Language Models
CAKE benchmark shows MCQ accuracy on cloud architecture plateaus near 99% above 3B parameters while free-response scores improve steadily with size, and reasoning steps help but tools hurt small models.
-
Phase-Associative Memory: Sequence Modeling in Complex Hilbert Space
PAM, a complex-valued associative memory model, exhibits steeper power-law scaling in loss and perplexity than a matched real-valued baseline when trained on WikiText-103 from 5M to 100M parameters.
-
DeepStack: Scalable and Accurate Design Space Exploration for Distributed 3D-Stacked AI Accelerators
DeepStack introduces a fast performance model and hierarchical search method for co-optimizing 3D DRAM stacking, interconnects, and distributed scheduling in AI accelerators, delivering up to 9.5x throughput gains over baselines.
-
FlashSAC: Fast and Stable Off-Policy Reinforcement Learning for High-Dimensional Robot Control
FlashSAC improves training speed and final performance of off-policy RL on high-dimensional robot tasks by reducing update frequency, increasing model scale, and bounding norms to limit critic error accumulation.
-
Data Scaling as Progressive Coverage of a Predictive Contribution Spectrum
Empirical study across 12 corpora shows tail slope of global-KL predictive contribution spectrum correlates with scaling exponents and effective truncation rank K(N) scales linearly with log N (R² 0.96 raw), supporting spectrum coverage as the driver of excess loss reduction.
-
The Compression Gap: Why Discrete Tokenization Limits Vision-Language-Action Model Scaling
Discrete action tokenization in VLA models creates an information bottleneck that prevents vision encoder scaling from improving performance, unlike continuous policies, as validated on the LIBERO benchmark.
-
Composer Vector: Style-steering Symbolic Music Generation in a Latent Space
Composer Vector steers symbolic music generation models in latent space at inference time to control and blend composer styles without retraining.
-
Revealing the Learning Dynamics of Long-Context Continual Pre-training
Industrial-scale LLMs require over 150B tokens for long-context continual pre-training to reach intrinsic saturation, with perplexity and retrieval-head attention providing stronger signals than needle-in-a-haystack tests.
-
Lifting Unlabeled Internet-level Data for 3D Scene Understanding
Unlabeled web videos processed by designed data engines generate effective training data that yields strong zero-shot and finetuned performance on 3D detection, segmentation, VQA, and navigation.
-
Spectral Edge Dynamics: An Analytical-Empirical Study of Phase Transitions in Neural Network Training
Spectral gaps in the Gram matrix of parameter updates control phase transitions such as grokking in neural network training.
-
Rethinking Language Model Scaling under Transferable Hypersphere Optimization
HyperP transfers optimal learning rates across model width, depth, tokens, and MoE granularity under Frobenius-sphere constraints, delivering stable scaling and 1.58x efficiency gains.
-
CLIP-RD: Relative Distillation for Efficient CLIP Knowledge Distillation
CLIP-RD adds VRD for cross-modality distillation consistency and XRD for bidirectional cross-modal symmetry to align student embedding geometry more closely with the teacher, yielding a 0.8 percentage point gain over prior distillation methods.
-
Diffusion Model for Manifold Data: Score Decomposition, Curvature, and Statistical Complexity
Diffusion models on manifold-supported data admit score decompositions whose statistical rates are controlled by intrinsic dimension and curvature.
-
Hawkeye: Reproducing GPU-Level Non-Determinism
Hawkeye enables exact CPU reproduction of NVIDIA Tensor Core matrix multiplications via systematic characterization of arithmetic behaviors across multiple GPU architectures and precisions.
-
MobileLLM-Flash: Latency-Guided On-Device LLM Design for Industry Scale Deployment
MobileLLM-Flash creates 350M-1.4B parameter LLMs via latency-guided search and attention skipping, delivering up to 1.8x faster prefill and 1.6x faster decode on mobile CPUs with comparable or better quality.
-
LPC-SM: Local Predictive Coding and Sparse Memory for Long-Context Language Modeling
LPC-SM is a hybrid architecture separating local attention, persistent memory, predictive correction, and control with ONT for memory writes, showing loss reductions on 158M-parameter models up to 4096-token contexts.
-
Capacity-Aware Mixture Law Enables Efficient LLM Data Optimization
CAMEL is a scaling law capturing nonlinear model-size and mixture interactions to extrapolate optimal data mixtures for large LLMs from small-model experiments, reducing optimization cost by 50% and improving benchmarks by up to 3%.
-
Spectral Condition for $\mu$P under Width-Depth Scaling
A unified spectral condition for μP under width-depth scaling reveals a transition at k=1 vs k≥2 transformations per residual block and enables stable feature learning for practical architectures like Transformers.
-
Brain-OF: An Omnifunctional Foundation Model for fMRI, EEG and MEG
Brain-OF is a multimodal foundation model for fMRI, EEG and MEG using any-resolution sampling, DINT attention with sparse MoE, and masked temporal-frequency pretraining on ~40 datasets to achieve superior downstream performance.
-
veScale-FSDP: Flexible and High-Performance FSDP at Scale
veScale-FSDP uses RaggedShard and structure-aware planning to support block-wise quantization and non-element-wise optimizers while delivering 5-66% higher throughput and 16-30% lower memory than prior FSDP systems at massive scale.
-
World Action Models are Zero-shot Policies
DreamZero uses a 14B video diffusion model as a World Action Model to achieve over 2x better zero-shot generalization on real robots than state-of-the-art VLAs, real-time 7Hz closed-loop control, and cross-embodiment transfer with 10-30 minutes of data.
-
Why Code, Why Now: An Information-Theoretic Perspective on the Limits of Machine Learning
Task information structure determines ML scaling success, with code's dense verifiable signals enabling predictable progress while sparse-feedback tasks like typical RL do not.
-
The Quantization Trap: Breaking Linear Scaling Laws in Multi-Hop Reasoning
Reducing precision from 16-bit to 8/4-bit in multi-hop reasoning creates a quantization trap that raises net energy consumption and degrades accuracy, breaking linear scaling laws.
-
When Less is More: The LLM Scaling Paradox in Context Compression
Larger LLM compressors in lossy setups often yield less faithful context reconstructions due to knowledge overwriting and semantic drift, with mid-sized models outperforming larger ones across 27 tested configurations.
-
Protein Autoregressive Modeling via Multiscale Structure Generation
PAR is a multi-scale autoregressive transformer framework for protein backbone generation that uses coarse-to-fine prediction, noisy context learning, and flow-based decoding to achieve high-quality unconditional and zero-shot conditional outputs.
-
Sparsity is Combinatorial Depth: Quantifying MoE Expressivity via Tropical Geometry
MoE Top-k routing equals the k-th elementary symmetric tropical polynomial, making sparsity combinatorial depth that scales capacity by binom(N,k) and gives MoE combinatorial resilience on manifolds.
-
The Hot Mess of AI: How Does Misalignment Scale With Model Intelligence and Task Complexity?
AI model failures on complex tasks become increasingly incoherent with longer reasoning chains, making consistent misalignment less likely than chaotic errors as capabilities scale.
-
AGZO: Activation-Guided Zeroth-Order Optimization for LLM Fine-Tuning
AGZO restricts ZO perturbations to an activation-derived low-rank subspace, claiming higher gradient cosine similarity and better benchmark performance than isotropic ZO baselines on Qwen3 and Pangu models.
-
CLARE: Continual Learning for Vision-Language-Action Models via Autonomous Adapter Routing and Expansion
CLARE is an exemplar-free continual learning framework for VLAs that autonomously expands modular adapters based on feature similarity and uses autoencoder routing for label-free deployment.
-
ALL-FEM: Agentic Large Language models Fine-tuned for Finite Element Methods
ALL-FEM fine-tunes LLMs on a corpus of verified FEniCS scripts and uses multi-agent workflows to automate finite element code generation, achieving 71.79% success on 39 benchmarks across elasticity, flow, and coupled problems.
-
On Harnessing Idle Compute at the Edge for Foundation Model Training
Cleave trains foundation models on heterogeneous edge devices by decomposing GEMM operations to exploit downlink-uplink asymmetry, achieving cloud-comparable speed and scaling to thousands of devices with fast failure recovery.
-
BOOST: BOttleneck-Optimized Scalable Training Framework for Low-Rank Large Language Models
BOOST delivers 1.46-2.27x end-to-end speedups for low-rank bottleneck LLMs by redesigning tensor parallelism around the bottleneck structure plus supporting optimizations.
-
TRINITY: An Evolved LLM Coordinator
A compact 0.6B-parameter coordinator with a 10K-parameter head uses evolutionary strategy to dynamically delegate roles to LLMs, achieving SOTA results such as 86.2% on LiveCodeBench.
-
SimScale: Learning to Drive via Real-World Simulation at Scale
SimScale synthesizes unseen driving states from real logs via neural rendering and reactive environments, generates pseudo-expert trajectories, and shows that co-training on real plus simulated data improves planning robustness and generalization on real benchmarks, with gains scaling by simulation
-
WiFo-2: a generalist foundation model unifies heterogeneous wireless system design
WiFo-2 is a space-time-frequency foundation model pretrained on heterogeneous CSI data that delivers strong zero-shot and few-shot performance across wireless communications and sensing tasks.
-
SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control
Scaling motion tracking models along size, data volume, and compute produces a foundation model for natural, robust humanoid whole-body control with downstream uses in kinematic planning and vision-language-action models.
-
Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail
Alpamayo-R1 introduces a VLA model with a Chain of Causation dataset and multi-stage SFT-plus-RL training that reports 12% better planning accuracy and 35% fewer close encounters versus trajectory-only baselines in driving tasks.
-
Scaling Laws Meet Model Architecture: Toward Inference-Efficient LLMs
A conditional scaling law fitted on over 200 models from 80M to 3B parameters identifies architectures that deliver up to 2.1% higher accuracy and 42% higher inference throughput than LLaMA-3.2 under the same training budget.
-
Foundation Models for Discovery and Exploration in Chemical Space
MIST models up to 10x larger than prior work, fine-tuned on over 400 structure-property tasks, match or exceed SOTA on benchmarks and demonstrate zero-shot olfactory perception mapping consistent with hyperbolic geometry.
-
A solution to generalized learning from small training sets found in infant repeated visual experiences of individual objects
Infant daily visual experiences of objects are dominated by repeated instances of few exemplars in lumpy similarity clusters, enabling category generalization from small training sets in computational models.
-
DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Driving
DriveVLA-W0 adds world modeling to predict future images in VLA models, overcoming sparse action supervision and amplifying data scaling laws on NAVSIM benchmarks and a large in-house dataset.
-
LightReasoner: Can Small Language Models Teach Large Language Models Reasoning?
LightReasoner distills supervision signals from SLM-LLM behavioral divergence to improve LLM reasoning on math benchmarks with up to 28.1% accuracy gains and 90-99% reductions in resources.
-
Adaptive Memory Momentum via a Model-Based Framework for Deep Learning Optimization
Presents a model-based proximal framework for adaptive momentum in first-order optimizers by using a two-plane approximation of the objective to dynamically set the memory coefficient online.
-
Don't Pass@k: A Bayesian Framework for Large Language Model Evaluation
A Dirichlet-prior Bayesian estimator for model success probability replaces Pass@k, delivering faster-converging and more stable rankings with credible intervals on math benchmarks.
-
GRACE-MoE: Grouping and Replication with Locality-Aware Routing for Efficient Distributed MoE Inference
GRACE-MoE integrates expert grouping, dynamic replication, and locality-aware routing with hierarchical sparse communication to reduce end-to-end latency in distributed SMoE inference.
-
Model Merging Scaling Laws in Large Language Models
Empirical scaling laws for LLM merging show a size-dependent floor and 1/k-like tail in cross-entropy loss that holds across architectures and merging methods.
-
CR-Net: Scaling Parameter-Efficient Training with Cross-Layer Low-Rank Structure
CR-Net uses cross-layer low-rank residuals in a dual-path network plus specialized recomputation to outperform prior low-rank methods on 60M-7B model pre-training while using less compute and memory.
-
HyperAdapt: Simple High-Rank Adaptation
HyperAdapt performs parameter-efficient fine-tuning by row- and column-wise diagonal scaling to induce high-rank updates with only n+m trainable parameters.