Ion Stoica

Continuous scoring from LLM-as-a-Verifier leads to state-of-the-art verification accuracy and improved sample efficiency in reinforcement learning tasks.

Existing benchmarks miss the mark on faithfulness, but a new dependency-aware checklist reveals the true performance gaps in T2I models.

DualEval reveals that unifying static and preference-based evaluations can lead to more reliable model rankings and deeper insights into item performance.

Playful learning strategies enable robots to acquire skills that boost performance on new tasks by over 20%, transforming how we approach robotic skill acquisition.

Evolving coding problems can restore meaningful evaluation metrics for frontier models, revealing their true capabilities and enabling self-improvement.

Forget hand-crafted environments: ClawEnvKit lets you automatically generate diverse, verified environments for claw-like agents from natural language, slashing costs by 13,800x.

Get up to 10% more throughput on your LLM disaggregation workloads just by swapping in this drop-in collective communications library with built-in compression.

Cut LLM cold starts from minutes to seconds by pre-materializing CUDA graph execution contexts, sidestepping brittle kernel patching and heavyweight checkpointing.

Scaling prompt learning by 17x without sacrificing accuracy is now possible, unlocking efficient self-improvement for LLM agents.

LLM-driven program evolution gets a smart upgrade: AdaEvolve dynamically allocates resources to promising solution candidates, leaving static schedules in the dust.

LLMs can now design GPU kernels that outperform both human experts and prior automated methods, thanks to a co-evolving world model that guides the search process.