Inference & Efficiency
A
Showing 1–30 of 127
-
Build a Validated Kubernetes Runtime for NVIDIA DynamoNVIDIA details how to build a validated Kubernetes runtime for DynamoTeams building AI applications need a repeatable path from a model configuration to a working inference endpoint. NVIDIA walks through building a validated Kubernetes runtime for Dynamo, its distributed inference framework, aiming to standardize how a chosen model setup becomes a running inference service and to cut operational effort.
-
Run AI Models for Edge Application Development with Just a Few CommandsNVIDIA shows how to run AI models for edge app development in a few commandsAn NVIDIA developer blog post explains how to run AI models for edge application development with just a few commands. It frames the topic around AI expanding beyond the data center, as developers build agents that act at the edge, such as monitoring driver attention and cabin conditions.
-
Flide: Accelerating Decision Inference with Exact State Forking
-
Latent Sparse Attention: Towards Subquadratic Ultra-Long Context Attention
-
A Better Critic Matters for Distribution Matching Distillation
-
DecisionRank: On-Policy Distillation for Decision-Model Rerankers
-
ACCT: Asymmetric Cross-LLM Context Transfer
-
Subquadratic Attention with a Hierarchical Tree-Structured Indexer
-
RAMA: Retrieval-Augmented Multi-Agent Routing with Episodic Memory for LEO Satellite Networks
-
The Knowledge Layer Model (KLM): Evidence-Visible AI Output, Routing Diagnostics, and Negative Results
-
There Is No Free Loop: When Does Looping a Trained Model Pay for Its Inference Compute?
-
Asynchronous Flow Map Language Models
-
When Does Test-Time Search Help World Action Models? An Empirical Audit of Candidate Selection and Simulator Integrity
-
Prune Gradually, Adapt Sparsely: Efficient Pruning of Large Language Models at Extreme Sparsity
-
Completeness and Cost of Retrieval Structures for Multi-Tool Requests in LLM Agents
-
How Much Can Tool Calls Reveal About You? Privacy Leakage from Tool-Call Traces
-
Text as an Index for Zero-Shot Cross-Domain Knowledge Tracing
-
Dynamical Memory Transformer: Selective Recurrent Memories with Constant Cost Attention
-
Trainable Vector Quantization for Large Language Models
-
ElTab: Elastically Looped Tabular Transformers
-
Stepwise-Querying: Distilling Reasoning without Access to Teacher Chain-of-Thought
-
From Training Exposure to Membership Evidence: A Controlled Study of OLMo 2
-
Precision--Recall Trade-off on the Inference Level for LLMs in Grammatical Error Correction
-
When Steering Shifts Discourse: Identifying and Intervening on Discourse-Level Changes due to Activation Steering in LLMs
-
KV-Cache Patching: Tracing How and Where Fine-Tuned Behaviors Arise
-
What We Keep of What We Forget: Replacement Operators and Routing Semantics in Training-Free MoE Expert Pruning
-
Meta Ray-Banで向かいの席から撮影→指の動きからスマホのパスワードを解読する盗み見攻撃Smart-glasses video of finger movements can crack phone passwordsTel Aviv University researchers presented 'Watch and Crack,' a paper showing that passwords can be inferred from finger movements recorded by smart glasses. The shoulder-surfing attack films from across the table with Meta Ray-Ban glasses to read smartphone passwords.
-
Rethinking Neural Thickets Without Weight Perturbations
-
An ODE-based Sampler for Discrete Diffusion Models
-
Difficulty Calibration Across Time for Membership Inference in Diffusion Models