# Neural architectures for resolving references in program code

Source: [Neural architectures for resolving references in program code](<https://www.sota2.com/research/paper/69e06d347d0172140ec7cc60>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-17T05:00:13.532Z

Authors: Gerg\\H{o} Szalay, Gergely Zsolt Kov\\'acs, S\\'andor Teleki, Bal\\'azs Pint\\'er, Tibor Gregorics

Published: 2026-04-15T16:44:56.000Z

## Abstract

Resolving and rewriting references is fundamental in programming languages. Motivated by a real-world decompilation task, we abstract reference rewriting into the problems of direct and indirect indexing by permutation. We create synthetic benchmarks for these tasks and show that well-known sequence-to-sequence machine learning architectures are struggling on these benchmarks. We introduce new sequence-to-sequence architectures for both problems. Our measurements show that our architectures outperform the baselines in both robustness and scalability: our models can handle examples that are ten times longer compared to the best baseline. We measure the impact of our architecture in the real-world task of decompiling switch statements, which has an indexing subtask. According to our measurements, the extended model decreases the error rate by 42%. Multiple ablation studies show that all components of our architectures are essential.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Indirect permutation problem on PI1-10](<https://www.sota2.com/research/sota/indirect-permutation-problem-on-pi1-10.md>): Indirect permutation problem; PI1-10; TA: 99.99.

- [Indirect permutation problem on PI10](<https://www.sota2.com/research/sota/indirect-permutation-problem-on-pi10.md>): Indirect permutation problem; PI10; TA: 100.

- [direct permutation problem on PD20](<https://www.sota2.com/research/sota/direct-permutation-problem-on-pd20.md>): direct permutation problem; PD20; TA: 99.95.

- [Indirect permutation problem on PI\_DICT (test)](<https://www.sota2.com/research/sota/indirect-permutation-problem-on-pi_dict-test.md>): Indirect permutation problem; PI\_DICT (test); TA: 99.94.

- [Indirect permutation problem on PI20](<https://www.sota2.com/research/sota/indirect-permutation-problem-on-pi20.md>): Indirect permutation problem; PI20; TA: 99.99.

- [direct permutation problem on PD40 (test)](<https://www.sota2.com/research/sota/direct-permutation-problem-on-pd40-test.md>): direct permutation problem; PD40 (test); TA: 99.99.

- [direct permutation problem on PD100 (test)](<https://www.sota2.com/research/sota/direct-permutation-problem-on-pd100-test.md>): direct permutation problem; PD100 (test); TA: 99.99.

- [Indirect permutation problem on PI40](<https://www.sota2.com/research/sota/indirect-permutation-problem-on-pi40.md>): Indirect permutation problem; PI40; TA: 100.

- [Indirect permutation problem on PI100](<https://www.sota2.com/research/sota/indirect-permutation-problem-on-pi100.md>): Indirect permutation problem; PI100; TA: 99.96.

- [direct permutation problem on PD1 10](<https://www.sota2.com/research/sota/direct-permutation-problem-on-pd1-10.md>): direct permutation problem; PD1 10; TA: 99.99.

- [direct permutation problem on PD10](<https://www.sota2.com/research/sota/direct-permutation-problem-on-pd10.md>): direct permutation problem; PD10; TA: 100.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.14073>)

[PDF](<https://arxiv.org/pdf/2604.14073.pdf>)
