AI Paper Atlas

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

Bridging Functional Correctness and Runtime Efficiency Gaps in LLM-Based Code Translation

Longhui Zhang, Jiahao Wang, Chenhao Hu, Bingyu Liang, Jing Li, Min Zhang

ICML 2026 · Large language models › Agents, code and math · approach: Execution-based code verification · 0 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 efficiency-oriented prompting (Correctness+Efficiency) instead of Correctness-OnlyQwen3-Next-80B / GPT-5 · CodeNet+F2SBench+SwiftBenchCA (%) and ET (ms): GPT-5 CA 86.4 -> 79.8, ET 528 -> 453Mixed
use SwiftTrans on Qwen2.5-7B instead of F2STransQwen2.5-7B · CodeNet+F2SBench+SwiftBenchCA (%): 84.6 -> 90.2Improved
use SwiftTrans instead of F2STrans on Qwen2.5-7BQwen2.5-7B · CodeNet+F2SBench+SwiftBenchET (ms): 731 -> 292Improved
increase demonstrations k from 0 to 1 (ablation)MpTranslator parallel ICL · SwiftBenchCA / ET: +1.3% CA, -142 ms ETImproved
increase demonstrations k from 1 to 3 (ablation)MpTranslator parallel ICL · SwiftBenchCA / ET: +0.7% CA, -18 ms ETImproved
increase acceleration layers n from 0 to 4 (ablation)Hierarchical guidance · SwiftBenchET / CA: -425 ms ET, -0.4% CAMixed
train with tgt^0 only supervision instead of hierarchical guidance (ablation)MpTranslator · SwiftBenchET (ms): ET = 860 msHurt
train with tgt^n only supervision (ablation)MpTranslator · SwiftBenchCA (%): CA = 85.1%Hurt
use bubble selection instead of all-pair selection (ablation)DiffSelector · SwiftBenchCA / ET: CA 89.1 -> 88.9, ET 439 -> 448No clear effect
add ordinal guidance loss L_og (ablation)DiffSelector · SwiftBenchCA (%): 86.1 -> 88.9Improved
add diff information to judge input (ablation)DiffSelector · SwiftBenchOrder Sensitivity (%): 27.5 -> 6.4Improved
add bi-judge training (ablation)DiffSelector · SwiftBenchOrder Sensitivity (%): 18.7 -> 6.4Improved

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