Ion Stoica

EditBench: Evaluating LLM Abilities to Perform Real-World Instructed Code Edits

Computer Agent Arena: Toward Human-Centric Evaluation and Analysis of Computer-Use Agents

SonicMoE: Accelerating MoE with IO and Tile-aware Optimizations

GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning

lmgame-Bench: How Good are LLMs at Playing Games?

DeepScholarBench: A Live Benchmark and Automated Evaluation for Generative Research Synthesis

LLMSELECTOR: Towards Model Selection Optimization for Compound AI Systems

vAttention: Verified Sparse Attention via Sampling

SLA: Beyond Sparsity in Diffusion Transformers via Fine-Tunable Sparse–Linear Attention

SPECS: Faster Test-Time Scaling through Speculative Drafts and Dynamic Switching

DeepScaleR: Effective RL Scaling of Reasoning Models via Iterative Context Lengthening

SciPro Arena: a Case Study of AI Agent Capabilities in Scientific Analysis Tasks

BARE: Leveraging Base Language Models for Few-Shot Synthetic Data Generation

Shepherd: Pattern-Guided Trajectory Selection for Coding Agents on SWE-Bench

R2E-Gym: Procedural Environment Generation and Hybrid Verifiers for Scaling Open-Weights SWE Agents

HashAttention: Semantic Sparsity for Faster Inference

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

Radial Attention: $\mathcal O(n \log n)$ Sparse Attention for Long Video Generation

Sparse VideoGen2: Accelerate Video Generation with Sparse Attention via Semantic-Aware Permutation

Faster Video Diffusion with Trainable Sparse Attention

Efficiently Scaling LLM Reasoning Programs with Certaindex

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

Fast Video Generation with Sliding Tile Attention

GameArena: Evaluating LLM Reasoning through Live Computer Games

JudgeBench: A Benchmark for Evaluating LLM-Based Judges

MPC-Minimized Secure LLM Inference

RouteLLM: Learning to Route LLMs from Preference Data

How to Evaluate Reward Models for RLHF

LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code

Exploring and Mitigating Adversarial Manipulation of Voting-Based Leaderboards

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

Sparse Video-Gen: Accelerating Video Diffusion Transformers with Spatial-Temporal Sparsity

Bench-O-Matic: Automating Benchmark Curation from Crowdsourced Data

Prompt-to-Leaderboard: Prompt-Adaptive LLM Evaluations

Efficient-vDiT: Efficient Video Diffusion Transformers With Attention Tile

A Statistical Framework for Ranking LLM-based Chatbots

Copilot Arena: A Platform for Code LLM Evaluation in the Wild

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

Post-Training Sparse Attention with Double Sparsity

The Berkeley Function Calling Leaderboard (BFCL): From Tool Use to Agentic Evaluation of Large Language Models

Test-Time RAG: Enhancing Long Context Understanding in LLMs with Retrieval-Augmented Mechanisms