# Towards Feedback-to-Plan Decisions for Self-Evolving LLM Agents in CUDA Kernel Generation

arXiv: https://arxiv.org/abs/2605.26720 · HTML: https://mlsanity.com/paper/2605.26720/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in General LLM agent frameworks and benchmarks, one of 16 directions in LLM agents and tool use.
- Builds on 28 of its 57 references that are in the map.
- First posted on arXiv 26 May 2026.

## Placed in

- LLM agents and tool use / [General LLM agent frameworks and benchmarks](https://mlsanity.com/agents/general-llm-agent-frameworks-and-benchmarks/)
- Synthetic data and self-improvement for LLMs / [Self-evolving LLM agents](https://mlsanity.com/synth/self-evolving-llm-agents/)

## Builds on (most cited first, 28 of 28)

- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [Towards Best Practices of Activation Patching in Language Models: Metrics and Methods](https://mlsanity.com/paper/2309.16042/) (2023)
- [Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents](https://mlsanity.com/paper/2505.22954/) (2025)
- [Evolution of Heuristics: Towards Efficient Automatic Algorithm Design Using Large Language Model](https://mlsanity.com/paper/2401.02051/) (2024)
- [KernelBench: Can LLMs Write Efficient GPU Kernels?](https://mlsanity.com/paper/2502.10517/) (2025)
- [ShinkaEvolve: Towards Open-Ended And Sample-Efficient Program Evolution](https://mlsanity.com/paper/2509.19349/) (2025)
- [Monte Carlo Tree Search for Comprehensive Exploration in LLM-Based Automatic Heuristic Design](https://mlsanity.com/paper/2501.08603/) (2025)
- [Propagation Kernels](https://mlsanity.com/paper/1410.3314/) (2014)
- [CUDA-L1: Improving CUDA Optimization via Contrastive Reinforcement Learning](https://mlsanity.com/paper/2507.14111/) (2025)
- [Towards Robust Agentic CUDA Kernel Benchmarking, Verification, and Optimization](https://mlsanity.com/paper/2509.14279/) (2025)
- [GraKeL: A Graph Kernel Library in Python](https://mlsanity.com/paper/1806.02193/) (2018)
- [Tiramisu: A Polyhedral Compiler for Expressing Fast and Portable Code](https://mlsanity.com/paper/1804.10694/) (2018)
- [Astra: A Multi-Agent System for GPU Kernel Performance Optimization](https://mlsanity.com/paper/2509.07506/) (2025)
- [Reducing Cost of LLM Agents with Trajectory Reduction](https://mlsanity.com/paper/2509.23586/) (2025)
- [CudaForge: An Agent Framework with Hardware Feedback for CUDA Kernel Optimization](https://mlsanity.com/paper/2511.01884/) (2025)
- [LLM4AD: A Platform for Algorithm Design with Large Language Model](https://mlsanity.com/paper/2412.17287/) (2024)
- [STARK: Strategic Team of Agents for Refining Kernels](https://mlsanity.com/paper/2510.16996/) (2025)
- [FlashInfer-Bench: Building the Virtuous Cycle for AI-driven LLM Systems](https://mlsanity.com/paper/2601.00227/) (2026)
- [Interpreting Emergent Planning in Model-Free Reinforcement Learning](https://mlsanity.com/paper/2504.01871/) (2025)
- [ConCuR: Conciseness Makes State-of-the-Art Kernel Generation](https://mlsanity.com/paper/2510.07356/) (2025)
- [SwizzlePerf: Hardware-Aware LLMs for GPU Kernel Performance Optimization](https://mlsanity.com/paper/2508.20258/) (2025)
- [PRAGMA: A Profiling-Reasoned Multi-Agent Framework for Automatic Kernel Optimization](https://mlsanity.com/paper/2511.06345/) (2025)
- [Prompt Valuation Based on Shapley Values](https://mlsanity.com/paper/2312.15395/) (2023)
- [Understanding and Optimizing Agentic Workflows via Shapley value](https://mlsanity.com/paper/2502.00510/) (2025)
- [AgentMonitor: A Plug-and-Play Framework for Predictive and Secure Multi-Agent Systems](https://mlsanity.com/paper/2408.14972/) (2024)
- [cuPilot: A Strategy-Coordinated Multi-agent Framework for CUDA Kernel Evolution](https://mlsanity.com/paper/2512.16465/) (2025)
- [Agentic Auto-Scheduling: An Experimental Study of LLM-Guided Loop Optimization](https://mlsanity.com/paper/2511.00592/) (2025)
- [Optimizing PyTorch Inference with LLM-Based Multi-Agent Systems](https://mlsanity.com/paper/2511.16964/) (2025)

## Abstract (arXiv)

Large language models (LLMs) have shown strong empirical gains as self-evolving agents for CUDA kernel generation, driven by feedback-conditioned planning across generations. However, how planning decisions attribute and combine heterogeneous feedback signals remains opaque. Standard end-to-end ablations fail to resolve this question, as iterative planning amplifies early perturbations and conflates feedback effects with trajectory-dependent drift. We introduce textttCUDAnalyst, a unified analysis layer for controlled, generation-level attribution of planning decisions to feedback components via trajectory freezing and selective feedback injection. textttCUDAnalyst enables stable generation-level evaluation and principled coalitional-style attribution of feedback effects and interactions. Our results show that explicit planning is beneficial only when feedback is aligned, that effective planning emerges from structured multi-feedback interactions, and that high-level plans from stronger reasoning models can partially transfer to weaker ones. These trends hold across reference backbones, representative workloads, and reference induction regimes, indicating that the identified feedback-to-plan structure is robust within the controlled axes studied.