ForgeKernel: Self-Evolving Kernels Catching and Surpassing Experts and Agents
cs.DC
Submitted: 2026-11-03
Updated: 2026-11-03
Code: https://github.com/OpenBMB/ForgeKernel
Terminology
Sources
- Kevin: Multi-Turn RL for Generating CUDA Kernels
- K-Search: LLM Kernel Generation via Co-Evolving Intrinsic World Model
- JustGoal: With Only a Goal, the AI Baseline Is Above the Expert Kernels
- DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model
- STARK: Strategic Team of Agents for Refining Kernels
- EGG: An Expert-Guided Agent Framework for Kernel Generation
- Kimi Linear: An Expressive, Efficient Attention Architecture
- CUDA-L1: Improving CUDA Optimization via Contrastive Reinforcement Learning
- CuBridge: An LLM-Based Framework for Understanding and Reconstructing High-Performance Attention Kernels
- AlphaEvolve: A coding agent for scientific and algorithmic discovery
- FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision
- KernelClaw: A Multi-Agent Framework for GPU Kernel Optimization with Dual-Memory Mechanisms
- Geak: Introducing Triton Kernel AI Agent & Evaluation Benchmarks
- Astra: A Multi-Agent System for GPU Kernel Performance Optimization
- Mirage: A Multi-Level Superoptimizer for Tensor Programs
- Gated Delta Networks: Improving Mamba2 with Delta Rule
- FlashInfer: Efficient and Customizable Attention Engine for LLM Inference Serving
- CAKE: Compiler-Agent Co-Design for Frontier Kernel Evolution
Related papers
- Programmable Probabilistic Computer with 1,000,000 p-bits
- BatchGen: An Architecture for Scalable and Efficient Batch Inference
- Seamless Reconfiguration for DAG-BFT (Technical Report)
- ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving
- iScheduler: Reinforcement Learning-Driven Continual Optimization for Large-Scale Resource Investment Problems
- TIRx: A Unified Tile Primitive ML Compiler Across Heterogeneous Backends