Subscribe
Research / page 12

Research

470 articles

Articles, page 12

221-240 of 470

  1. Gradient Low-Rank Projection (GaLore): How Dynamic Subspaces Enable Full-Parameter LLM Pre-Training with Low-Rank Memory

    Pre-training foundation large language models has historically required massive compute clusters, largely due to the memory footprint of optimizer states rather than the model weights themselves. While parameter-efficient fine-tuning methods such as Low-Rank Adaptation (LoRA) freeze weights and introduce small adapter matrices, they fail when applied to pre-training from scratch because they restrict parameter updates to a fixed, static low-rank manifold. Gradient Low-Rank Projection (GaLore),

    AI
    8 min read

  2. Microscaling Formats in Production LLM Inference: NVFP4 vs. MXFP4 Architecture, Block Quantization, Kernel Scaling, and Perplexity Degradation

    As frontier language models scale into hundreds of billions of parameters, memory bandwidth and VRAM capacity remain the primary bottlenecks in production inference serving. While 8-bit floating-point formats (FP8 E4M3 and E5M2) achieved mainstream adoption across Hopper and Ada generation hardware, the industry is transitioning to 4-bit floating-point (FP4) arithmetic to double inference throughput and halve memory footprints. Directly quantizing deep transformer weights and activations into 4

    AI
    7 min read

  3. Variational Autoencoders: Mathematical Derivation of the ELBO, the Reparameterization Trick, and Mitigating Posterior Collapse

    Variational Autoencoders: Mathematical Derivation of the ELBO, the Reparameterization Trick, and Mitigating Posterior Collapse Traditional autoencoders map high-dimensional data into deterministic latent vectors. While effective for dimensionality reduction and non-linear feature compression, deterministic autoencoders fail as generative models because their latent representations lack continuous probabilistic structure. Unregularized latent spaces contain wide regions of empty space and severe

    AI
    11 min read

  4. Non-Contrastive Representation Learning: How Barlow Twins and VICReg Prevent Feature Collapse via Redundancy Reduction

    Self-supervised representation learning provides the foundation for modern foundation models across computer vision, audio, and multimodal systems. By training deep neural networks to produce compact vector embeddings without human annotations, self-supervised pre-training enables models to capture rich semantic structures directly from raw data. Historically, the dominant approach to self-supervised learning was contrastive learning, popularized by architectures such as SimCLR (Chen et al., 20

    edu
    9 min read

  5. The Platonic Representation Hypothesis: How Foundation Models Across Modalities Converge on a Shared Geometry of Reality

    In 375 BC, Plato introduced the Allegory of the Cave, describing prisoners who could only perceive reality through shadows projected onto a stone wall. In modern machine learning, deep neural networks ingest distinct sensory shadows of the physical world: vision models process pixel arrays, large language models ingest tokenized text, and speech models parse raw acoustic waveforms. Historically, the machine learning literature assumed that models trained on different modalities, objectives, and

    AI
    6 min read

  6. Predictive KV Cache Warming in Production LLM Serving: Architecture, Session Prefetching, and TTFT Latency Shaving

    Predictive KV Cache Warming in Production LLM Serving: Architecture, Session Prefetching, and TTFT Latency Shaving In long-context large language model (LLM) deployments, Time-to-First-Token (TTFT) represents the primary latency bottleneck. As input prompts scale from 8,000 tokens to 128,000 tokens in multi-turn coding agents, legal document analysis, and enterprise retrieval pipelines, prefill execution consumes between 85% and 95% of total GPU compute time per request. Standard optimization

    AI
    5 min read

  7. Autonomous Retail AI Agent Luna Fires Employee Following Context Retrieval Breakdown and Human Intervention

    In an empirical field deployment examining autonomous AI workforce management, research firm Andon Labs reported that its storefront manager agent, Luna, decided to fire a human retail employee after months of operational infractions. The incident, which unfolded at the Andon Market retail location in San Francisco, represents one of the first documented instances of an autonomous large language model agent managing physical store operations and executing a personnel termination decision. Luna,

    Agents
    3 min read

  8. Vector Quantization and VQ-VAEs: How Discrete Codebooks, Straight-Through Estimators, and Commitment Losses Power Multimodal Tokenization

    Autoregressive sequence models excel at discrete token prediction. In natural language processing, words and subwords map onto categorical vocabularies where token identity is exact and cross-entropy loss provides direct likelihood optimization. Continuous multi-dimensional signals—such as images, video frames, raw audio waveforms, and robotic sensorimotor trajectories—present a fundamental mismatch for standard transformer architectures. Historically, variational autoencoders (VAEs) bridged ra

    AI
    7 min read

  9. Federated LLM Fine-Tuning in Production: FedLoRA, Differential Privacy, and Cross-Silo Aggregation Architectures

    Fine-tuning foundation large language models on proprietary data is standard enterprise practice, but centralizing sensitive tokens into a single data lake is frequently prohibited. Regulatory frameworks such as HIPAA in healthcare, GDPR and Article 10 of the EU AI Act in Europe, and regional data residency mandates across APAC and North America prevent cross-border or cross-institutional data aggregation. Federated Learning (FL) resolves this bottleneck by decoupling model training from data c

    AI
    8 min read

  10. The QK and OV Circuits in Transformers: How Bilinear Attention Routing and Subspace Projections Move Information

    The QK and OV Circuits in Transformers: How Bilinear Attention Routing and Subspace Projections Move Information The standard mathematical presentation of multi-head self-attention, introduced in Vaswani et al. (2017), describes the layer as a sequence of matrix projections followed by scaled dot-product operations, concatenation, and an output projection. While computationally efficient for parallel GPU hardware, this formulation obscures the fundamental linear mechanics governing how transfor

    AI
    9 min read

  11. Agent Skills in Production: Progressive Disclosure, Sandboxed Execution, and Procedural Memory Scaffolding

    Autonomous AI agents deployed in enterprise environments face an operational bottleneck: general-purpose frontier models possess broad linguistic reasoning, but lack the domain-specific procedural discipline required to complete multi-step workflows reliably. When engineering teams attempt to bridge this gap, standard techniques encounter severe architectural ceilings: 1. Monolithic system prompts degrade reasoning performance as instructions accumulate, triggering attention saturation, needle

    AI
    7 min read

  12. Rejection Sampling Fine-Tuning in Large Language Models: How Best-of-N Filtering, Reward Oracles, and Distillation Align Neural Policies

    Rejection Sampling Fine-Tuning: How Filtering Model Outputs by Reward Optimizes Alignment Without Policy Gradient Instability Post-training alignment has become a defining phase in modern large language model development. While supervised fine-tuning (SFT) teaches a model to follow instructions and adopt structured formats, aligning model behavior with human preferences, safety criteria, and domain accuracy requires optimizing against reward signals. Historically, this optimization has been ap

    AI
    10 min read

  13. Discrete Diffusion in Large Language Models: How Continuous-Time Markov Chains, Absorbing States, and Score Entropy Challenge Autoregressive Generation

    The dominance of autoregressive architectures in large language models rests on a fundamental mathematical formulation: the chain rule of probability. By factoring the joint distribution of a sequence into a product of conditional probabilities, $p(x) = \prod_{i=1}^N p(x_i \mid x_{<i})$, autoregressive models reduce text generation to sequential next-token prediction. While this left-to-right causal factorization has scaled effectively across compute regimes, it imposes rigid operational constr

    AI
    8 min read

  14. Continuous Pre-Training in Production: Domain Adaptation, Replay Buffers, Learning Rate Restarts, and Catastrophic Forgetting Mitigation

    Continuous Pre-Training in Production: Domain Adaptation, Replay Buffers, Learning Rate Restarts, and Catastrophic Forgetting Mitigation Adapting general-purpose foundation models to specialized enterprise domains (such as clinical medicine, corporate law, quantitative finance, and proprietary software codebases) presents a fundamental architectural challenge. While Retrieval-Augmented Generation (RAG) and Supervised Fine-Tuning (SFT) remain standard first-line approaches, both exhibit severe s

    LLM
    7 min read

  15. Hybrid SSM-Transformer Architectures: How Interleaving Attention and Recurrence Solves the State-Retrieval Trade-Off

    Hybrid SSM-Transformer Architectures: How Interleaving Attention and Recurrence Solves the State-Retrieval Trade-Off Autoregressive language models face a fundamental tension between inference efficiency and long-context retrieval capacity. Pure Transformer architectures scale quadratic computational complexity during sequence prefill and linear key-value (KV) cache memory consumption during autoregressive token generation. Conversely, pure State Space Models (SSMs) and linear recurrent neural

    AI
    8 min read

  16. Study: Why Labor-Saving LLMs Incline Scientists to Do More Work Less Well

    A theoretical study published by researchers from Princeton University, the University of Washington, and collaborating institutions models how large language models alter researchers' time allocation across projects. The authors find that by reducing time friction across different stages of the research lifecycle, AI assistants increase the opportunity cost of researcher time, creating economic incentives to publish a higher volume of less thoroughly refined papers. The paper, titled The unint

    AI
    2 min read

  17. Sparse Attention and BigBird: How Window, Global, and Random Graphs Preserve Turing Completeness in Linear Time

    Standard self-attention in transformer architectures scales quadratically with sequence length. Computing full pairwise interactions between n tokens requires evaluating an n x n attention matrix, yielding O(n^2) computational complexity and memory consumption. While hardware accelerators and IO-aware tiling algorithms like FlashAttention optimize memory traffic, the quadratic compute and KV footprint remains a barrier for processing long contexts, document-level summarization, and genomic seque

    AI
    7 min read

  18. Low-Precision Quantization Kernels in Production: Comparing Marlin, ExLlamaV2, FlashInfer, and BitBLAS

    Low-Precision Quantization Kernels in Production: Comparing Marlin, ExLlamaV2, FlashInfer, and BitBLAS Architecture, Memory Bandwidth, and Decoding Throughput Autoregressive large language model (LLM) serving operates under two distinct compute regimes: a compute-bound prefill phase and a memory-bandwidth-bound decode phase. While processing the initial prompt involves matrix-matrix multiplications (GEMM) with high arithmetic intensity, generating tokens one by one requires matrix-vector multip

    AI
    9 min read

  19. xLSTM: How Exponential Gating and Matrix Memory Scale Recurrent Neural Networks

    xLSTM: How Exponential Gating and Matrix Memory Scale Recurrent Neural Networks For over two decades following its introduction by Hochreiter and Schmidhuber (1997), the Long Short-Term Memory (LSTM) network served as the dominant architecture for sequence modeling. By introducing the constant error carousel and multiplicative gating, LSTMs mitigated the vanishing gradient problem that plagued vanilla recurrent neural networks. However, the emergence of the Transformer architecture (Vaswani et

    AI
    10 min read

  20. Identity Preference Optimization (IPO): How Exact Loss Inversion Prevents Overfitting in Direct Alignment

    The post-training alignment of large language models underwent a structural shift with the introduction of Direct Preference Optimization (Rafailov et al., 2023). By reparameterizing the closed-form optimal policy under a Kullback-Leibler (KL) constrained Reinforcement Learning from Human Feedback (RLHF) objective, DPO eliminated the need to fit an explicit reward model or maintain complex actor-critic rollout loops. However, standard DPO introduces a distinct mathematical pathology: under dete

    AI
    7 min read