2023

Evaluating and Explaining Large Language Models for Code Using Syntactic Structures

Palacio, David N, Velasco, Alejandro, Rodriguez-Cardenas, Daniel et al.

Understand

Large Language Models (LLMs) for code are a family of high-parameter, transformer-based neural networks pre-trained on massive datasets of both natural and programming languages.

These models are rapidly being employed in commercial AI-based developer tools, such as GitHub CoPilot.
However, measuring and explaining their effectiveness on programming tasks is a challenging proposition, given their size and complexity.
The methods for evaluating and explaining LLMs for code are inextricably linked.

Built on

M. Lewis, Y. Liu, N. Goyal et al. , “BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension.” [Online]. Available: http://arxiv.org/abs/1910.13461
Original
1910
Earlier work this paper cites.
A. Kanade, P. Maniatis, G. Balakrishnan, and K. Shi, “Learning and evaluating contextual embedding of source code.” [Online]. Available: http://arxiv.org/abs/2001.00059
Original
2001
Earlier work this paper cites.
Z. Feng, D. Guo, D. Tang et al. , “CodeBERT: A pre-trained model for programming and natural languages.” [Online]. Available: http://arxiv.org/abs/2002.08155
Original
2002
Earlier work this paper cites.
Y. Bengio, R. Ducharme, and P. Vincent, “A neural probabilistic language model,” Advances in Neural Information Processing Systems , vol. 3, pp. 1137–1155, 2003
2003
Earlier work this paper cites.
J. E. Hopcroft, R. Motwani, and J. D. Ullman, Introduction to Automata Theory, Languages, and Computation (3rd Edition) . USA: Addison-Wesley Longman Publishing Co., Inc., 2006
2006
Earlier work this paper cites.
S. Ren, D. Guo, S. Lu et al. , “CodeBLEU: a Method for Automatic Evaluation of Code Synthesis,” vol. 1949, no. Weaver 1955, 2020, arXiv: 2009.10297. [Online]. Available: http://arxiv.org/abs/2009.10297
Original
2009
Earlier work this paper cites.
D. Guo, S. Ren, S. Lu et al. , “GraphCodeBERT: Pre-training code representations with data flow.” [Online]. Available: http://arxiv.org/abs/2009.08366
Original
2009
Earlier work this paper cites.
V. Belle and I. Papantonis, “Principles and Practice of Explainable Machine Learning,” 2020, arXiv: 2009.11698. [Online]. Available: http://arxiv.org/abs/2009.11698
Original
2009
Earlier work this paper cites.
C. Watson, N. Cooper, D. Nader-Palacio et al. , “A systematic literature review on the use of deep learning in software engineering research,” CoRR , vol. abs/2009.06520, 2020. [Online]. Available: https://arxiv.org/abs/2009.06520
Original
2009
Earlier work this paper cites.
J. Pearl, Causality: models, reasoning, and inference , 2009
2009
Earlier work this paper cites.
C. Molnar, G. Casalicchio, and B. Bischl, “Interpretable Machine Learning – A Brief History, State-of-the-Art and Challenges,” no. 01, pp. 1–15, 2020, arXiv: 2010.09337. [Online]. Available: http://arxiv.org/abs/2010.09337
Original
2010
Earlier work this paper cites.
V. Raychev, M. T. Vechev, and E. Yahav, “Code completion with statistical language models,” Proceedings of the 35th ACM SIGPLAN Conference on Programming Language Design and Implementation , 2014
2014
Earlier work this paper cites.
A. Karpathy, J. Johnson, and F. Li, “Visualizing and understanding recurrent networks,” CoRR , vol. abs/1506.02078, 2015. [Online]. Available: http://arxiv.org/abs/1506.02078
Original
2015
Earlier work this paper cites.
J. Pearl, M. Glymour, and N. P.Jewell, Causal Inference in Statistics, A Primer , 2016
2016
Earlier work this paper cites.
A. Vaswani, N. Shazeer, N. Parmar et al. , “Attention is all you need,” in Proceedings of the 31st International Conference on Neural Information Processing Systems , ser. NeurIPs’17. Red Hook, NY, USA: Curran Associates Inc., 2017, p. 6000–6010
2017
Earlier work this paper cites.
B. Kim, M. Wattenberg, J. Gilmer et al. , “Interpretability beyond feature attribution: Quantitative Testing with Concept Activation Vectors (TCAV),” 35th International Conference on Machine Learning, ICML 2018 , vol. 6, pp. 4186–4195, 2018
2018
Earlier work this paper cites.
M. Allamanis, M. Brockschmidt, and M. Khademi, “Learning to represent programs with graphs,” in International Conference on Learning Representations , 2018. [Online]. Available: https://openreview.net/forum?id=BJOFETxR-
2018
Earlier work this paper cites.
J. Pearl and D. Mackenzie, The book of why: The New Science of Cause and Effect , 2018
2018
Earlier work this paper cites.

Similar

Z. Chen, S. J. Kommrusch, M. Tufano et al. , “Sequencer: Sequence-to-sequence learning for end-to-end program repair,” IEEE Transactions on Software Engineering , pp. 1–1, 2019
2019
Cited alongside, same era.
H. Husain, H.-H. Wu, T. Gazit et al. , “CodeSearchNet challenge: Evaluating the state of semantic code search,” arXiv preprint arXiv:1909.09436 , 2019
Original
2019
Cited alongside, same era.
C. Molnar, Interpretable Machine Learning , 2019, https://christophm.github.io/interpretable-ml-book/
2019
Cited alongside, same era.
M. Allamanis, “The adverse effects of code duplication in machine learning models of code,” in Onward! OOPLSA 2019 , 2019, pp. 143–153. [Online]. Available: https://doi.org/10.1145/3359591.3359735
2019
Cited alongside, same era.
R. M. Karampatsis and C. Sutton, “Maybe deep neural networks are the best choice for modeling source code,” arXiv , no. Lm, 2019
2019
Cited alongside, same era.
A. Paszke, S. Gross, F. Massa et al. , “Pytorch: An imperative style, high-performance deep learning library,” in Advances in Neural Information Processing Systems 32 , H. Wallach, H. Larochelle, A. Beygelzimer et al. , Eds. Curran Associates, Inc., 2019, pp. 8024–8035. [Online]. Available: http://papers.neurips.cc/paper/9015-pytorch-an-imperative-style-high-performance-deep-learning-library.pdf
2019
Cited alongside, same era.
C. Watson, M. Tufano, K. Moran et al. , “On learning meaningful assert statements for unit test cases,” in Proceedings of the ACM/IEEE 42nd International Conference on Software Engineering , ser. ICSE ’20. New York, NY, USA: Association for Computing Machinery, 2020, p. 1398–1409. [Online]. Available: https://doi.org/10.1145/3377811.3380429
2020
Cited alongside, same era.
R. M. Karampatsis, H. Babii, R. Robbes et al. , “Open-Vocabulary Models for Source Code (Extended Abstract),” Proceedings - 2020 ACM/IEEE 42nd International Conference on Software Engineering: Companion, ICSE-Companion 2020 , pp. 294–295, 2020
2020
Cited alongside, same era.
L. Gao, S. Biderman, S. Black et al. , “The pile: An 800gb dataset of diverse text for language modeling,” 2020
2020
Cited alongside, same era.
T. Wolf, L. Debut, V. Sanh et al. , “Transformers: State-of-the-art natural language processing,” in Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations . Online: Association for Computational Linguistics, Oct. 2020, pp. 38–45. [Online]. Available: https://www.aclweb.org/anthology/2020.emnlp-demos.6
2020
Cited alongside, same era.
M. Ciniselli, N. Cooper, L. Pascarella et al. , “An empirical study on the usage of BERT models for code completion,” CoRR , vol. abs/2103.07115, 2021. [Online]. Available: https://arxiv.org/abs/2103.07115
Original
2021
Cited alongside, same era.
M. Chen, J. Tworek, H. Jun et al. , “Evaluating Large Language Models Trained on Code,” Jul. 2021, arXiv:2107.03374 [cs]. [Online]. Available: http://arxiv.org/abs/2107.03374
Original
2021
Cited alongside, same era.

Then

M. Chen, J. Tworek, H. Jun et al. , “Evaluating Large Language Models Trained on Code,” 2021. [Online]. Available: http://arxiv.org/abs/2107.03374
Original
2021
Later among the works it cites.
A. Sharma, V. Syrgkanis, C. Zhang, and E. Kıcıman, “DoWhy : Addressing Challenges in Expressing and Validating Causal Assumptions,” 2021
2021
Later among the works it cites.
S. Lu, D. Guo, S. Ren et al. , “Codexglue: A machine learning benchmark dataset for code understanding and generation,” CoRR , vol. abs/2102.04664, 2021
Original
2021
Later among the works it cites.
Y. Wan, W. Zhao, H. Zhang et al. , “What Do They Capture? – A Structural Analysis of Pre-Trained Language Models for Source Code,” Feb. 2022, arXiv:2202.06840 [cs]. [Online]. Available: http://arxiv.org/abs/2202.06840
Original
2022
Later among the works it cites.
S. Black, S. Biderman, E. Hallahan et al. , “Gpt-neox-20b: An open-source autoregressive language model,” 2022. [Online]. Available: https://arxiv.org/abs/2204.06745
Original
2022
Later among the works it cites.
F. F. Xu, U. Alon, G. Neubig, and V. J. Hellendoorn, “A Systematic Evaluation of Large Language Models of Code,” May 2022, arXiv:2202.13169 [cs]. [Online]. Available: http://arxiv.org/abs/2202.13169
Original
2022
Later among the works it cites.
S. Troshin and N. Chirkova, “Probing Pretrained Models of Source Code,” Nov. 2022, arXiv:2202.08975 [cs]. [Online]. Available: http://arxiv.org/abs/2202.08975
Original
2022
Later among the works it cites.
J. A. H. López, M. Weyssow, J. S. Cuadrado, and H. Sahraoui, “AST-Probe: Recovering abstract syntax trees from hidden representations of pre-trained language models,” Sep. 2022, arXiv:2206.11719 [cs]. [Online]. Available: http://arxiv.org/abs/2206.11719
Original
2022
Later among the works it cites.
Anonymous, “Interpretingcodegeneration,” https://anonymous.4open.science/r/InterpretingCodeGeneration-75E4/ , 2022
2022
Later among the works it cites.
E. Nijkamp, B. Pang, H. Hayashi et al. , “A conversational paradigm for program synthesis,” arXiv preprint , 2022
2022
Later among the works it cites.
Replit, “Replit Ghostwriter,” https://replit.com/site/ghostwriter , May 2023
2023
Closest in time.
J. Liu, C. S. Xia, Y. Wang, and L. Zhang, “Is your code generated by chatgpt really correct? rigorous evaluation of large language models for code generation,” 2023
2023
Closest in time.
Y. Liu, C. Tantithamthavorn, Y. Liu, and L. Li, “On the Reliability and Explainability of Automated Code Generation Approaches,” Feb. 2023, arXiv:2302.09587 [cs]. [Online]. Available: http://arxiv.org/abs/2302.09587
Original
2023
Closest in time.
W. X. Zhao, K. Zhou, J. Li et al. , “A Survey of Large Language Models,” Apr. 2023, arXiv:2303.18223 [cs]. [Online]. Available: http://arxiv.org/abs/2303.18223
Original
2023
Closest in time.
H. Vasconcelos, G. Bansal, A. Fourney et al. , “Generation Probabilities Are Not Enough: Exploring the Effectiveness of Uncertainty Highlighting in AI-Powered Code Completions,” Feb. 2023, arXiv:2302.07248 [cs]. [Online]. Available: http://arxiv.org/abs/2302.07248
Original
2023
Closest in time.
M. Brunsfeld, A. Hlynskyi, P. Thomson et al. , “tree-sitter/tree-sitter: v0.20.8,” Apr. 2023. [Online]. Available: https://doi.org/10.5281/zenodo.7798573
2023
Closest in time.
D. N. Palacio, N. Cooper, A. Rodriguez et al. , “Toward a theory of causation for interpreting neural code models,” 2023
2023
Closest in time.
E. Nijkamp, B. Pang, H. Hayashi et al. , “Codegen: An open large language model for code with multi-turn program synthesis,” 2023
2023
Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…

Reading the bibliography…