Ion Stoica

papers

9,870

total citations

papers (21)

From Crowdsourced Data to High-quality Benchmarks: Arena-Hard and Benchbuilder Pipeline

ICML 2025arXiv

357

citations

LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset

ICLR 2024arXiv

352

citations

Break the Sequential Dependency of LLM Inference Using Lookahead Decoding

ICML 2024arXiv

257

citations

OR-Bench: An Over-Refusal Benchmark for Large Language Models

ICML 2025arXiv

104

citations

Online Speculative Decoding

ICML 2024arXiv

citations

Accelerating Quadratic Optimization with Reinforcement Learning

NEURIPS 2021arXiv

citations

LLM-Assisted Code Cleaning For Training Accurate Code Generators

ICLR 2024arXiv

citations

MuxServe: Flexible Spatial-Temporal Multiplexing for Multiple LLM Serving

ICML 2024arXiv

citations

WorldModelBench: Judging Video Generation Models As World Models

NEURIPS 2025arXiv

citations

RLlib Flow: Distributed Reinforcement Learning is a Dataflow Problem

NEURIPS 2021arXiv

citations

GameArena: Evaluating LLM Reasoning through Live Computer Games

ICLR 2025arXiv

citations

Trustless Audits without Revealing Data or Models

ICML 2024arXiv

citations

Establishing Best Practices in Building Rigorous Agentic Benchmarks

NEURIPS 2025arXiv

citations

VisionArena: 230k Real World User-VLM Conversations with Preference Labels

CVPR 2025arXiv

citations

Twilight: Adaptive Attention Sparsity with Hierarchical Top-$p$ Pruning

NEURIPS 2025arXiv

citations

Context-Aware Streaming Perception in Dynamic Environments

ECCV 2022arXiv

citations

HashAttention: Semantic Sparsity for Faster Inference

ICML 2025arXiv

citations

R2E: Turning any Github Repository into a Programming Agent Environment

ICML 2024

citations

Ion Stoica

papers (21)

Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena

Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference

Representing Long-Range Context for Graph Neural Networks with Global Attention

From Crowdsourced Data to High-quality Benchmarks: Arena-Hard and Benchbuilder Pipeline

LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset

Break the Sequential Dependency of LLM Inference Using Lookahead Decoding

OR-Bench: An Over-Refusal Benchmark for Large Language Models

Online Speculative Decoding

Accelerating Quadratic Optimization with Reinforcement Learning

LLM-Assisted Code Cleaning For Training Accurate Code Generators

MuxServe: Flexible Spatial-Temporal Multiplexing for Multiple LLM Serving

WorldModelBench: Judging Video Generation Models As World Models

RLlib Flow: Distributed Reinforcement Learning is a Dataflow Problem

GameArena: Evaluating LLM Reasoning through Live Computer Games

Trustless Audits without Revealing Data or Models

Establishing Best Practices in Building Rigorous Agentic Benchmarks

VisionArena: 230k Real World User-VLM Conversations with Preference Labels

Twilight: Adaptive Attention Sparsity with Hierarchical Top-$p$ Pruning

Context-Aware Streaming Perception in Dynamic Environments

HashAttention: Semantic Sparsity for Faster Inference

R2E: Turning any Github Repository into a Programming Agent Environment

papers (21)

Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena

Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference

Representing Long-Range Context for Graph Neural Networks with Global Attention

From Crowdsourced Data to High-quality Benchmarks: Arena-Hard and Benchbuilder Pipeline

LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset

Break the Sequential Dependency of LLM Inference Using Lookahead Decoding

OR-Bench: An Over-Refusal Benchmark for Large Language Models

Online Speculative Decoding

Accelerating Quadratic Optimization with Reinforcement Learning

LLM-Assisted Code Cleaning For Training Accurate Code Generators

MuxServe: Flexible Spatial-Temporal Multiplexing for Multiple LLM Serving

WorldModelBench: Judging Video Generation Models As World Models

RLlib Flow: Distributed Reinforcement Learning is a Dataflow Problem

GameArena: Evaluating LLM Reasoning through Live Computer Games

Trustless Audits without Revealing Data or Models

Establishing Best Practices in Building Rigorous Agentic Benchmarks

VisionArena: 230k Real World User-VLM Conversations with Preference Labels

Twilight: Adaptive Attention Sparsity with Hierarchical Top-$p$ Pruning

Context-Aware Streaming Perception in Dynamic Environments

HashAttention: Semantic Sparsity for Faster Inference

R2E: Turning any Github Repository into a Programming Agent Environment