vLLM Semantic Router
Signal-driven decision routing for mixture-of-modality deployments.
16 papers / 2025-2026
Selected Works
LLM Routing, Research, and AI Infra.
Papers and infrastructure work around routing as a systems problem.
Paper Archive
Papers on routing, systems, and inference optimization.
A research archive spanning semantic routing, agent behavior, and infrastructure efficiency.
Token-Budget-Aware Pool Routing for Cost-Efficient LLM Inference
Proposes token-budget-aware pool routing that estimates each request's token budget online and dispatches it to short- or long-context serving pools, reducing GPU cost while improving stability for LLM inference.
vLLM Semantic Router: Signal Driven Decision Routing for Mixture-of-Modality Models
Introduces vLLM Semantic Router as a signal-driven routing framework for mixture-of-modality deployments, composing heterogeneous signals into deployment-specific policies across cost, privacy, latency, and safety.
The Workload-Router-Pool Architecture for LLM Inference Optimization: A Vision Paper from the vLLM Semantic Router Project
Synthesizes recent routing, fleet, multimodal, and governance work into the Workload-Router-Pool architecture, framing routing as one layer in a broader inference optimization stack.
Visual Confused Deputy: Exploiting and Defending Perception Failures in Computer-Using Agents
Formalizes the visual confused deputy as a security failure mode for computer-use agents and proposes a dual-channel guardrail for validating click targets and action reasoning before execution.
Outcome-Aware Tool Selection for Semantic Routers: Latency-Constrained Learning Without LLM Inference
Presents OATS, an offline embedding-refinement method that improves semantic-router tool ranking under single-digit millisecond CPU budgets without adding serving-time model inference.
Adaptive Vision-Language Model Routing for Computer Use Agents
Proposes Adaptive VLM Routing to estimate step difficulty in computer-use agents and route each action to the cheapest model that can still satisfy a target reliability threshold.
98x Faster LLM Routing Without a Dedicated GPU: Flash Attention, Prompt Compression, and Near-Streaming for the vLLM Semantic Router
Combines flash attention, prompt compression, and near-streaming body processing to cut routing latency from seconds to tens of milliseconds on lightweight shared serving hardware.
inference-fleet-sim: A Queueing-Theory-Grounded Fleet Capacity Planner for LLM Inference
Introduces a queueing-theory-grounded fleet planner and discrete-event simulator for sizing multi-pool LLM GPU fleets against P99 TTFT targets without requiring up-front hardware profiling.
FleetOpt: Analytical Fleet Provisioning for LLM Inference with Compress-and-Route as Implementation Mechanism
Derives the minimum-cost two-pool LLM fleet directly from workload distributions and P99 TTFT targets, then maps the optimal boundary to a deployable compress-and-route strategy.
The 1/W Law: An Analytical Study of Context-Length Routing Topology and GPU Generation Gains for LLM Inference Energy Efficiency
Derives the 1/W law, showing that tokens per watt roughly halve whenever the serving context window doubles, making context-length routing a first-order energy-efficiency lever.
Conflict-Free Policy Languages for Probabilistic ML Predicates: A Framework and Case Study with the Semantic Router DSL
Shows how probabilistic ML predicates in policy languages can silently co-fire on the same query and adds conflict detection plus a softmax prevention mechanism in the Semantic Router DSL.
From Inference Routing to Agent Orchestration: Declarative Policy Compilation with Cross-Layer Verification
Extends the Semantic Router DSL from stateless per-request routing to multi-step agent workflows, compiling verified decision artifacts across orchestration, Kubernetes, and protocol layers.
Knowledge Access Beats Model Size: Memory Augmented Routing for Persistent AI Agents
Shows that conversational memory and retrieval-grounded routing let a lightweight 8B model recover most of a much larger model's performance on persistent user-specific queries while dramatically reducing cost.
Fast and Faithful: Real-Time Verification for Long-Document Retrieval-Augmented Generation Systems
Adds a real-time verification layer for long-document RAG systems, handling contexts up to 32K tokens while balancing latency with grounding coverage.
When to Reason: Semantic Router for vLLM
Routes prompts by reasoning requirements so reasoning is only invoked when it pays off, improving accuracy while cutting token usage and latency versus always-on reasoning.
Category-Aware Semantic Caching for Heterogeneous LLM Workloads
Proposes category-aware semantic caching with category-specific similarity thresholds, TTLs, and quotas, using a hybrid split between in-memory HNSW retrieval and external document storage.
Open Source
Infrastructure and standards work beyond the router.
Maintainer, steering, and reviewer roles across gateways, service mesh, and inference infrastructure.
Envoy Gateway
Manages Envoy Proxy as a standalone or Kubernetes-based application gateway.
Envoy AI Gateway
Manages unified access to generative AI services built on Envoy Gateway.
vLLM AIBrix
Cost-efficient and pluggable infrastructure components for GenAI inference.
Higress
AI gateway and AI-native API gateway.
Istio
Connects, secures, controls, and observes services.
Kiali
Observability console for Istio with service mesh.
Aeraki Mesh
Manages any layer-7 protocols in a service mesh.
Merbridge
Uses eBPF to speed up service mesh data paths.
Kubernetes Gateway API
Role-oriented, portable, and expressive interfaces for Kubernetes networking.
Kubernetes Ingress2Gateway
Converts Ingress resources to Gateway API resources.