Papers
Topics
Authors
Recent
Detailed Answer
Quick Answer
Concise responses based on abstracts only
Detailed Answer
Well-researched responses based on abstracts and relevant paper content.
Custom Instructions Pro
Preferences or requirements that you'd like Emergent Mind to consider when generating responses
Gemini 2.5 Flash
Gemini 2.5 Flash 63 tok/s
Gemini 2.5 Pro 49 tok/s Pro
GPT-5 Medium 14 tok/s Pro
GPT-5 High 19 tok/s Pro
GPT-4o 100 tok/s Pro
Kimi K2 174 tok/s Pro
GPT OSS 120B 472 tok/s Pro
Claude Sonnet 4 37 tok/s Pro
2000 character limit reached

RobustFill: Neural Program Learning under Noisy I/O (1703.07469v1)

Published 21 Mar 2017 in cs.AI

Abstract: The problem of automatically generating a computer program from some specification has been studied since the early days of AI. Recently, two competing approaches for automatic program learning have received significant attention: (1) neural program synthesis, where a neural network is conditioned on input/output (I/O) examples and learns to generate a program, and (2) neural program induction, where a neural network generates new outputs directly using a latent program representation. Here, for the first time, we directly compare both approaches on a large-scale, real-world learning task. We additionally contrast to rule-based program synthesis, which uses hand-crafted semantics to guide the program generation. Our neural models use a modified attention RNN to allow encoding of variable-sized sets of I/O pairs. Our best synthesis model achieves 92% accuracy on a real-world test set, compared to the 34% accuracy of the previous best neural synthesis approach. The synthesis model also outperforms a comparable induction model on this task, but we more importantly demonstrate that the strength of each approach is highly dependent on the evaluation metric and end-user application. Finally, we show that we can train our neural models to remain very robust to the type of noise expected in real-world data (e.g., typos), while a highly-engineered rule-based system fails entirely.

Citations (375)
List To Do Tasks Checklist Streamline Icon: https://streamlinehq.com

Collections

Sign up for free to add this paper to one or more collections.

Summary

  • The paper introduces a neural program synthesis model that generates programs from noisy I/O examples, achieving a 92% accuracy rate.
  • It employs a dual-attention RNN architecture with dynamic programming constraints to robustly handle variable-sized I/O pairs in string transformations.
  • The study contrasts synthesis with induction models, demonstrating that the synthesis approach sustains 80% accuracy under noise compared to traditional rule-based systems.

Overview of RobustFill: Neural Program Learning under Noisy I/O

The paper "RobustFill: Neural Program Learning under Noisy I/O" undertakes a comprehensive paper of two paradigms in neural program learning: neural program synthesis and induction. The researchers compare these approaches head-to-head on a real-world dataset, prioritizing robustness against noise. They introduce a novel attentional recurrent neural network (RNN) architecture designed to handle variable-sized sets of input/output (I/O) pairs. This architecture is evaluated using a Programming By Example (PBE) system for string transformations, much like Microsoft’s FlashFill. The authors claim a 92% accuracy with their synthesis model, a significant improvement over the prior best system's 34%.

Program Synthesis vs. Induction

In neural program synthesis, the network generates a program from I/O examples, while in neural program induction, it generates output directly without explicitly creating a program. Notably, the synthesis model can verify program consistency by executing candidate programs to check if they align with all observed I/O examples—a luxury not afforded to induction models. The synthesis approach's 92% accuracy is achieved by using a double attention mechanism and by leveraging Dynamic Programming (DP) constraints during beam search. The authors emphasize that each model type's effectiveness can vary significantly depending on the application context and evaluation metrics.

Key Contributions

  • Enhanced Encoding with Attentional Layers: The proposed architecture enables attention over both inputs and outputs, enhancing the model's ability to handle the complexity inherent in program synthesis tasks.
  • Adaptation to Real-World Conditions: The research involves testing against noise, such as typos in user-generated data, to gauge robustness. Under noise conditions, their neural model retains 80% accuracy, whereas rule-based systems falter dramatically with only 6% accuracy.
  • Comparative Analysis: The paper presents a distinct focus on contrasting the strengths of synthesis and induction under real-world test conditions.
  • Domain-Specific Language (DSL): A specialized DSL was developed, equipping the neural model with a rich set of string transformation abilities, essential for achieving high expressivity in the synthesized programs.

Numerical Results

The synthesis model, leveraging a dual-attentional encoder-decoder structure, demonstrated an accuracy of 92% on 205 instances of a real-world test set, signifying a decisive outmatch over previous benchmarks. This model's ability to maintain robust performance in noisy environments distinctly illustrates the potential of neural approaches over traditional rule-based systems.

Implications and Future Work

The findings underscore the potential of neural networks in situations involving noisy input data, highlighting their superior adaptability over rule-based systems. From a theoretical perspective, this research contributes to the ongoing discourse on latent vs. explicit program representations. Practically, it suggests pathways for more integrated program synthesis tools, particularly in user-centric applications like data wrangling in spreadsheets. Future research avenues could focus on exploring hybrid models that combine the strengths of synthesis and induction or expanding the DSL to cover more programming paradigms.

Conclusion

The investigation into RobustFill elucidates the transformative potential of advanced neural architectures in program synthesis. By offering a detailed comparative analysis, the authors provide critical insights that could guide future research and development in neural program learning. RobustFill’s demonstrated robustness to noise and its systematic evaluation against pre-existing systems firmly position it as a significant contribution to the AI field.

Dice Question Streamline Icon: https://streamlinehq.com

Follow-Up Questions

We haven't generated follow-up questions for this paper yet.

Don't miss out on important new AI/ML research

See which papers are being discussed right now on X, Reddit, and more:

“Emergent Mind helps me see which AI papers have caught fire online.”

Philip

Philip

Creator, AI Explained on YouTube