Patel, HarshHarshPatelVenkatesh, PraveenPraveenVenkateshSahni, ShivamShivamSahniJain, VarunVarunJainAnand, MrinalMrinalAnandSingh, MayankMayankSingh2025-08-312025-08-312022-01-08[9781450385824]10.1145/3493700.34937562-s2.0-85122670375https://d8.irins.org/handle/IITG2025/26202Computers are devices that execute precise instructions provided to them using various programming languages. However, the idea of delivering instructions to a computer through natural language could vastly simplify the act of programming as a specific task. Generating code from high-level descriptions for a given program is a significantly challenging task and has been an active area of research in the natural language processing domain. In this paper, we present a novel feedback-based deep learning approach for synthesizing code from human-specified descriptions. Inspired by the dual-learning mechanism, our framework uses a feedback loss to produce more consistent and robust predictions. We show how our approach fares empirically on standard code generation datasets and achieves state-of-the-art results on the NAPS (Natural Program Synthesis) dataset.falseConference Paper310-3118 January 20220cpConference Proceeding0