AI Paper Atlas

Research map / Large language models / Agents, code and math

From Large to Small: Transferring CUDA Optimization Expertise via Reasoning Graph

Junfeng Gong, Zhiyi Wei, Junying Chen, Cheng Liu, Huawei Li

ICLR 2026 · Large language models › Agents, code and math · approach: LLMs for optimization and heuristics · 10 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add reasoning graph retrieval with random search (ReGraphT) over RAGQwen2.5-Coder-7B-Instruct · CUDAEvalspeedup@n: 7.09 -> 12.89Improved
replace random-sampling traversal with Monte Carlo Graph Search (ablation)ReGraphT · CUDAEvalpass@n and speedup@n: +2.2% pass@n, +1.33% speedup@nImproved
replace random-sampling traversal with Monte Carlo Graph Search (ablation)ReGraphT · ParEvalpass@n and speedup@n: +1.7% pass@n, +0.34% speedup@nImproved
MCGS instead of random traversal on Qwen2.5-Coder-7B (ablation)ReGraphT · ParEvalpass@n: 51.7 -> 50.0Hurt
add chain-of-thought promptingQwen2.5-Coder-7B-Instruct · ParEvalpass@1: 38.3 -> 35.0Hurt
increase ReGraph size from 30 to 1000 samples (ablation)ReGraphT-MCGS · CUDAEvalpass@10: 61.3 -> 71.2Improved
increase ReGraph size beyond convergence (500 to 1000 samples) (ablation)ReGraphT-MCGS · CUDAEvalspeedup@10: 14.15 -> 14.21No clear effect
increase MCGS max attempts from 5 to 20 at budget 200 (ablation)ReGraph-MCGS · CUDAEvalspeedup@n: 11.99 -> 13.43Improved
MCGS instead of ReGraph traversal at budget 200 with 10 attempts (ablation)ReGraph · CUDAEvalspeedup@n: 11.30 -> 12.88Improved
use partial-credit reward instead of strict reward (ablation)ReGraphT-MCGS · CUDAEvalpass rate and speedupHurt
use rollout-based reward instead of strict reward (ablation)ReGraphT-MCGS · CUDAEvalpass rate and speedupNo clear effect
add ReGraphT-MCGS on hard-level tasksQwen2.5-Coder-7B-Instruct · CUDAEval hardpass@n: 43.1 -> 51.0Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy