Knowledge catalogue
arXiv:2605.10453v1 Announce Type: cross Abstract: Speculative decoding speeds up autoregressive generation in Large Language Models (LLMs) through a two-step procedure, where a lightweight draft model
arXiv:2605.10220v1 Announce Type: cross Abstract: The formation timescale of the Milky Way thick disk is one of the central debates in Galactic archaeology. The age-metallicity relation (AMR), formati
arXiv:2605.09311v1 Announce Type: cross Abstract: Unlike most static material properties widely studied in the machine learning literature, ionic transport properties are inherently dynamic, making th
arXiv:2605.08866v1 Announce Type: cross Abstract: Inverse optimization (IO) seeks to infer the parameters of a decision-maker's objective from observed context--action data. We study noiseless IO, whe
arXiv:2605.09281v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) models achieve remarkable performance by sparsely activating specialized experts, yet their massive parameters in experts pose
arXiv:2605.10344v1 Announce Type: new Abstract: Test-time scaling has become an effective paradigm for improving the reasoning ability of large language models by allocating additional computation dur
arXiv:2603.05301v2 Announce Type: replace Abstract: Inductive spatio-temporal kriging infers signals at unobserved locations from observed sensors, but real-world observations are often incomplete and
arXiv:2605.08830v1 Announce Type: cross Abstract: End-to-end autonomous driving requires models to understand traffic scenes, infer driving intent, and generate executable motion plans. Recent vision-
We published new research on how we serve post-trained Qwen3 235B models on NVIDIA GB200 NVL72 Blackwell racks. GB200 is a major step up over Hopper for high-throughput inference on large MoE models,
arXiv:2605.08234v1 Announce Type: cross Abstract: Long-context LLM inference is bottlenecked by the memory and bandwidth cost of reading large KV caches during decoding. KV compression reduces this co
This post analyzes the strategic directions and priorities of OpenAI, Anthropic, and xAI by examining their publicly listed job openings, inferring each company's focus areas, capabilities development
Enterprise AI adoption has crossed a threshold: The question is no longer whether to invest, but how to do it wisely. As agentic workloads multiply and inference costs rise, AI choice — the ability to
arXiv:2603.18636v2 Announce Type: replace Abstract: Diffusion Transformers (DiTs) achieve strong video generation quality but suffer from high inference cost due to dense 3D attention, motivating spar
Release b9102 adds support for the im2col_3d operator to the SYCL backend in llama.cpp, expanding GPU-accelerated tensor operations for Intel GPU inference. The im2col_3d operation converts 3D input v
B9103 is a recent release of llama.cpp , a C/C++ implementation for large language model inference. This release represents one of the latest versions in the project's active development cycle, buildi
arXiv:2605.07123v1 Announce Type: new Abstract: In-context reinforcement learning (ICRL) refers to the ability of RL agents to adapt to new tasks at inference time without parameter updates by conditi
arXiv:2605.06785v1 Announce Type: cross Abstract: Inference-time scaling methods rely on Process Reward Models (PRMs), which are often poorly calibrated and overestimate success probabilities. We prop
arXiv:2512.20974v3 Announce Type: replace-cross Abstract: Bayesian Reinforcement Learning (BRL), a subclass of Meta-Reinforcement Learning (Meta-RL), provides a principled framework for generalisation
arXiv:2509.21172v2 Announce Type: replace Abstract: Inverse reinforcement learning (IRL) aims to infer rewards from observed behavior, but rewards are not identified from the policy alone: many reward
arXiv:2602.02832v3 Announce Type: replace Abstract: Forecasting physical systems over long horizons from irregularly sampled observations demands models that are stable, computationally efficient, and
arXiv:2605.07315v1 Announce Type: new Abstract: Chain-of-thought (CoT) reasoning improves large language models (LLMs) on difficult tasks, but it also makes inference expensive because every intermedi
arXiv:2605.06676v1 Announce Type: cross Abstract: Long-context inference in Large Language Models (LLMs) is bottlenecked by the linear growth of Key-Value (KV) cache memory. Existing KV cache compress
arXiv:2605.06680v1 Announce Type: cross Abstract: Flow matching generates data by integrating a learned velocity field, where the number of integration steps (NFE) directly determines inference cost.
arXiv:2605.06758v1 Announce Type: cross Abstract: Relative spatial relations provide a compact representation of spatial structure and are fundamental to relative spatial reasoning in 3D layout genera
arXiv:2509.25584v2 Announce Type: replace Abstract: Vision-language models achieve incredible performance across a wide range of tasks, but their large size makes inference costly. Recent work has sho
arXiv:2509.05276v4 Announce Type: replace-cross Abstract: Mainstream Transformer-based large language models face major efficiency bottlenecks: training computation scales quadratically with sequence
arXiv:2605.08005v1 Announce Type: new Abstract: Test-Time Adaptation (TTA) aims to improve time series forecasting under distribution shifts by using limited observations revealed during inference. Ho
arXiv:2605.06726v1 Announce Type: new Abstract: Inferring the identity of wildlife species from daily movement data alone is a challenging task. We train sequence models on large-scale, 7-species GPS
arXiv:2605.07355v1 Announce Type: cross Abstract: Video-language models (VLMs) face rapid inference costs as visual token counts scale with video length. For example, 32 frames at 448{imes}448 resolut
arXiv:2605.08070v1 Announce Type: new Abstract: A standard technique for scaling inference-time reasoning is Self-Consistency, whereby multiple candidate answers are sampled from an LLM and the most c
Release b9094 is a commit/version of llama.cpp, the main playground for developing features for the GGML library. llama.cpp enables LLM inference in C/C++ , providing efficient tools for running large
b9095 is a release of llama.cpp, a project focused on LLM inference in C/C++ . This release likely includes updates, bug fixes, or feature improvements to the llama.cpp codebase, which enables efficie
Build b9101 of llama.cpp includes improvements to the AllReduce functionality for tensor-parallel GPU inference, introducing a NCCL-free implementation using CUDA kernels for cross-GPU communication.
The search results don't contain specific information about release b9087. Based on the context, b9087 is a release from llama.cpp, a C/C++ implementation for LLM inference . This release likely conta
b9073 is a release of llama.cpp, a C/C++ implementation for LLM inference . The release likely contains bug fixes, performance improvements, or new features for the llama.cpp project, which provides o
B9077 is a release of llama.cpp, a C/C++ library for LLM inference. The llama.cpp project follows a rapid release cycle with multiple releases published in a single day. This build version represents
B9082 is a release of llama.cpp, a C/C++ implementation for LLM inference . The release is available in multiple platform-specific binary distributions, including Windows CPU and ARM64 variants. This
llamacpp, a popular C++ inference engine for large language models, will soon support MTP (likely Media Transfer Protocol or a model-specific protocol), as announced by Clem Delangue. This addition wi
We’ve rolled out a significant update to Google Kubernetes Engine (GKE) that solves one of the most annoying problems in cloud infrastructure: cold start latency. GKE now has up to 4x faster node star
b9060 is a release build identifier for llama.cpp, an open-source LLM inference framework implemented in C/C++. This build represents a snapshot in the project's ongoing development, containing bug fi
Release b9063 is a build of llama.cpp, a project for LLM inference in C/C++. llama.cpp enables efficient large language model execution on consumer hardware through optimized implementations and quant
arXiv:2505.00020v2 Announce Type: replace Abstract: Using a legally obtained dataset of 34 copyrighted O'Reilly Media books, we apply the DE-COP membership inference attack method to investigate wheth
arXiv:2605.04980v1 Announce Type: cross Abstract: Activation-based steering provides control of LLM behavior at inference time, but the dominant paradigm reduces each concept to a single direction who
arXiv:2605.04507v1 Announce Type: new Abstract: Negotiation agents must infer what their counterpart values, update those beliefs over dialogue turns, and choose actions under uncertainty. End-to-end
arXiv:2605.04535v1 Announce Type: new Abstract: Inferring continuum models directly from video is hampered by two facts: the recorded field is uncalibrated image intensity rather than a physical state
arXiv:2605.04581v1 Announce Type: new Abstract: Light field (LF) image super-resolution benefits from Epipolar Plane Images (EPIs), whose line slopes explicitly encode disparity. However, existing Tra
arXiv:2605.05125v1 Announce Type: new Abstract: Target trial emulation (TTE) enables causal questions to be studied with observational data when randomized controlled trials (RCTs) are infeasible. Yet
arXiv:2510.17903v3 Announce Type: replace-cross Abstract: This paper tackles the challenging problem of jointly inferring time-varying network topologies and imputing missing data from partially obser
Post-training quantization is a technique that reduces model size and improves inference performance by converting weights and activations to lower precision formats after training is complete. NVIDIA
arXiv:2602.02282v2 Announce Type: replace Abstract: Inferring spatial transcriptomics (ST) from histology enables scalable histogenomic profiling, yet current methods are largely restricted to single-
arXiv:2605.04450v1 Announce Type: cross Abstract: Generative Recommender (GR) inference places embedding hot caches (EMB) and KV caches in direct competition for limited GPU HBM: allocating more memor
arXiv:2508.11196v2 Announce Type: replace Abstract: Recent advances in vision-language models (VLMs) have demonstrated strong generalization in natural image tasks. However, their performance often de
arXiv:2603.01097v2 Announce Type: replace Abstract: Continuous knowledge updating for pre-trained large language models (LLMs) is increasingly necessary yet remains challenging. Although inference-tim
arXiv:2602.10144v2 Announce Type: replace-cross Abstract: Minimizing the inference cost and latency of foundation models has become a crucial area of research. Optimization approaches include theoreti
llama.cpp b9041 is a release build of llama.cpp, an LLM inference library written in C/C++ . The build identifier indicates this is an intermediate development release from the ggml-org/llama.cpp proj
b9047 is a build release of llama.cpp, an open-source C/C++ library for running large language model inference on consumer hardware. The release includes pre-compiled binaries for multiple platforms i
arXiv:2605.02290v1 Announce Type: new Abstract: Distilling large reasoning models is essential for making Long-CoT reasoning practical, as full-scale inference remains computationally prohibitive. Exi
arXiv:2601.21684v2 Announce Type: replace Abstract: Test-Time Scaling enhances the reasoning capabilities of Large Language Models by allocating additional inference compute to broaden the exploration
arXiv:2505.19841v2 Announce Type: replace-cross Abstract: This work is focussed on the inversion task of inferring the distribution over parameters of interest leading to multiple sets of observations
arXiv:2510.09472v2 Announce Type: replace Abstract: Despite the remarkable progress in neural models, their ability to generalize, a cornerstone for applications such as logical reasoning, remains a c