Subscribe
Research / page 14

Research

470 articles

Articles, page 14

261-280 of 470

  1. Request Hedging in Production LLM Serving: Architecture, Tail-Latency Mitigation, and Cancellation Protocols

    In distributed computing, tail latency—the 95th, 99th, and 99.9th percentiles—dictates overall user experience, service-level agreements (SLAs), and multi-step agent execution reliability. While median response times (P50) in large language model (LLM) serving often appear acceptable, tail latencies frequently degrade by 4x to 10x. In multi-turn chat applications, real-time voice agents, and multi-agent DAG pipelines, a single straggler request stalls entire execution chains. Request hedging, a

    LLM
    8 min read

  2. LayerSkip and Self-Speculative Decoding: How Layer Dropout and Shared Early Exits Accelerate LLM Generation

    LayerSkip and Self-Speculative Decoding: How Layer Dropout and Shared Early Exits Accelerate LLM Generation Standard autoregressive large language model (LLM) inference is severely bottlenecked by memory bandwidth. In transformer decoders, generating a sequence of $N$ tokens requires loading all model parameters from High Bandwidth Memory (HBM) to on-chip SRAM $N$ separate times. While speculative decoding mitigates this bandwidth tax by using a smaller draft model to propose candidate tokens v

    AI
    8 min read

  3. Modern Hopfield Networks: How Continuous Energy Landscapes Explain Transformer Attention and Exponential Memory

    When Vaswani et al. introduced the Transformer architecture in 2017, scaled dot-product self-attention was presented primarily as a pragmatic computational mechanism: an efficient, highly parallelizable alternative to recurrence and convolutions. By computing pairwise inner products between queries and keys, normalizing via softmax, and taking a weighted sum of values, attention allowed models to route information dynamically across arbitrarily distant tokens. For several years, self-attention

    AI
    10 min read

  4. Fine-Grained Access Control in Enterprise RAG: Pre-Filtering vs. Post-Filtering, Zanzibar ReBAC Models, and Zero-Trust Retrieval Architecture

    Deploying Retrieval-Augmented Generation (RAG) across enterprise knowledge repositories introduces a security boundary that simple vector search was never designed to enforce. In corporate environments spanning Google Workspace, Microsoft SharePoint, Notion, Confluence, and internal ticket systems, access permissions are dynamic, hierarchical, and deeply nested. Attempting to enforce security at the prompt generation layer by instructing language models to ignore unauthorized context is fundame

    RAG
    6 min read

  5. Deep Double Descent: Why Overparameterization Defies the Classical Bias-Variance Trade-Off

    For decades, statistical learning theory rested on a foundational tenet: the bias-variance trade-off. According to classical machine learning textbooks, increasing model capacity reduces bias on the training set but inevitably inflates variance on unseen test data. The resulting risk curve forms a familiar U-shape: underfitting on the left, an optimal capacity in the center, and severe overfitting on the right. Modern deep learning and large language models (LLMs) fundamentally contradicted thi

    AI
    7 min read

  6. Demystifying Agent Skills: Empirical Study of 8,000+ Runs Shows Procedural Anchoring Beats Knowledge Injection

    A multi-institution study from researchers at Princeton University, UC San Diego, and collaborating labs provides the first large-scale empirical analysis of how "skills" (modular instruction packages loaded at inference time) alter autonomous AI agent trajectories. Analyzing 8,135 experimental trials across diverse model architectures, benchmarks, and agent harnesses, the authors establish that skills improve task completion primarily by acting as procedural anchors rather than by injecting mis

    AI
    2 min read

  7. Minimum Bayes Risk Decoding: How Utility Consensus and Distributional Reranking Outperform Mode-Seeking in Large Language Models

    Autoregressive large language models generate text by estimating conditional probability distributions over token sequences. In conventional inference pipelines, generating the final output sequence relies almost entirely on search algorithms designed to locate high-probability trajectories: greedy decoding, beam search, or stochastic sampling with temperature and top-p filtering. However, theoretical and empirical analyses demonstrate that the most probable sequence under a model's distributio

    AI
    7 min read

  8. The Curse of Multilinguality in Large Language Models: Capacity Dilution, Tokenizer Fertility, and Representation Interference

    The Curse of Multilinguality in Large Language Models: Capacity Dilution, Tokenizer Fertility, and Representation Interference Training a single transformer foundation model to process dozens or hundreds of languages is one of the central goals of modern natural language processing. In theory, massive multilingual pre-training unlocks positive cross-lingual transfer: low-resource languages gain syntactic, factual, and reasoning capabilities from the rich supervision available in high-resource l

    AI
    7 min read

  9. LLM Load Shedding and Overload Control in Production: Adaptive Admission, Preemption Economics, and Graceful Degradation

    LLM Load Shedding and Overload Control in Production: Adaptive Admission, Preemption Economics, and Graceful Degradation Standard web services rely on well-established overload protection patterns: reverse proxies monitor CPU utilization, memory thresholds, or static queue depths and reject excess HTTP requests with 429 Too Many Requests or 503 Service Unavailable status codes. When applied naively to Large Language Model (LLM) serving infrastructure, these conventional heuristics fail catastro

    LLM
    9 min read

  10. Reward Model Overoptimization in Large Language Models: How Goodhart's Law, Proxy Exploitation, and KL Drift Degrade Alignment

    Post-training alignment of large language models relies on optimizing a policy toward objectives defined by human intent and preferences. Because querying human evaluators during every step of continuous reinforcement learning or high-throughput rejection sampling is computationally and logistically infeasible, alignment workflows construct a parameterised proxy reward model. Trained on pairwise preference datasets through formulations such as the Bradley-Terry model, this proxy acts as a surrog

    AI
    6 min read

  11. Dynamic Few-Shot Example Selection in Production: Semantic Retrieval, Diversity Reranking, and Cache-Aligned Prompt Architectures

    In-context learning (ICL) remains one of the most practical mechanisms for steering large language models on specialized tasks, structured output parsing, domain-specific classification, and API tool calling. While zero-shot prompts rely entirely on the model's parametric memory, few-shot prompting provides concrete input-output demonstrations that anchor the model's generation trajectory. In enterprise production environments, however, static few-shot prompting quickly hits operational limits.

    AI
    9 min read

  12. Logit Soft-Capping in Large Language Models: How Tanh-Based Bounds Prevent Attention Saturation and Numerical Instability

    In modern autoregressive Transformer training, maintaining numerical stability across trillions of tokens requires strict control over intermediate activation magnitudes. As models grow deeper and wider, pre-softmax logits in self-attention mechanisms and final vocabulary projection layers frequently drift toward extreme values. When logit values grow unconstrained, standard softmax normalization enters a saturation regime where output probabilities collapse into near one-hot distributions, caus

    AI
    7 min read

  13. Mental World Modeling: Why Autonomous AI Agents Fail Without Simulating Human Beliefs

    Mental World Modeling: Why Autonomous AI Agents Fail Without Simulating Human Beliefs Current foundation world models, from video simulators like Sora and Genie to spatial representations like JEPA and Marble, focus almost exclusively on the physical mechanics of an environment. They track geometry, object positions, motion trajectories, and visual continuity. However, when autonomous agents interact with humans in collaborative, medical, or domestic settings, physical state tracking alone cons

    AI
    4 min read

  14. Reversible Transformers: How Invertible Residual Blocks Eliminate Activation Memory in Deep Networks

    Training deep transformer models is primarily bounded by activation memory rather than parameter storage. During the forward pass of standard backpropagation, automatic differentiation engines cache intermediate activations across every attention head, layer normalization, and feed-forward sublayer so they can be referenced during the backward pass to evaluate gradients. For a transformer with N layers, sequence length L, batch size B, and hidden dimension d_model, storing these activations requ

    AI
    6 min read

  15. Netflix Details GenRec LLM-Native Recommendation Architecture in Production A/B Trials

    Netflix has detailed GenRec, an internal machine learning architecture that adapts open-weight large language models for production recommendation ranking. The system replaces hand-crafted feature pipelines with natural-language context engineering, achieving measurable improvements in live A/B trials while reducing required training labels by up to 40 times. For years, industrial recommendation engines at scale have depended on complex feature stores tracking thousands of engineered numerical

    Netflix
    3 min read

  16. Rank Collapse in Deep Transformers: Why Pure Attention Degenerates Doubly Exponentially and How Skip Connections Preserve Capacity

    When the Transformer architecture was introduced in 2017 with the seminal paper "Attention Is All You Need", the central thesis was that recurrence and convolution could be completely discarded in favor of stacked self-attention mechanisms. However, theoretical analysis has shown that the title's premise is mathematically incomplete. Stacking pure self-attention layers in isolation does not produce an expressive deep model: it triggers a catastrophic failure mode known as rank collapse. In a fo

    AI
    7 min read

  17. Item Response Theory Audit of 192 LLMs Exposes Safety Benchmark Redundancies, Over-Refusal Distortions, and Sandbagging

    A psychometric evaluation of 192 frontier and open-weight language models across eight major safety benchmarks has revealed structural flaws in current safety testing methodologies. The research, conducted by Joshua Fonseca Rivera, Neil Shah, David Demitri Africa, and Konstantinos Voudouris with support from the UK AI Security Institute and the UK Department for Science, Innovation, and Technology (DSIT), applies Item Response Theory (IRT) to analyze 5,255 evaluation items. The findings demonst

    AI
    3 min read

  18. In-Context Learning as Implicit Gradient Descent: How Transformers Optimize Models in Forward Activations

    When large language models (LLMs) adapt to new tasks from a handful of prompt demonstrations, their static weights remain completely untouched. No backpropagation runs through the network, no optimizer updates parameters, and no gradients are calculated. Yet, the model’s predictions improve steadily as more input-output examples are added to the prompt context. For years, this phenomenon (known as in-context learning, or ICL) was treated as an empirical black box. Recent theoretical and mechani

    AI
    8 min read

  19. No Positional Embeddings (NoPE): How Causal Masking and Attention Geometry Encode Sequence Order

    A foundational tenet of the Transformer architecture established by Vaswani et al. (2017) is permutation equivariance. Because standard self-attention calculates token interactions purely through pairwise dot products across sets of vectors, shuffling the order of input tokens yields identical outputs up to the corresponding permutation. To establish word order, standard transformer models inject explicit positional information, ranging from learned absolute position embeddings (APE) to sinusoid

    AI
    6 min read

  20. Hugging Face ICML 2026 Audit: AI Coding Agents Falsify Claims Across 23% of 2,226 Examined Papers

    Hugging Face has published the findings of its ICML 2026 Open Reproductions challenge, a large-scale community audit that deployed autonomous AI coding agents to test the experimental claims of 2,226 accepted machine learning papers. The 19-day initiative involved 1,221 researchers and developers using tools including Claude Code, OpenAI Codex, Cursor, and OpenResearch orx. Participants generated 6,816 publicly auditable reproduction logbooks and executed 2,962 cloud compute jobs, examining rou

    AI
    2 min read