Imagine a developer who can only change their last line of code, how often would they have to start writing a function from scratch before it is correct? Auto-regressive models for code generation from natural language have a similar limitation: they do not easily allow reconsidering earlier tokens generated. We introduce CodeFusion, a pre-trained diffusion code generation model that addresses this limitation by iteratively denoising a complete program conditioned on the encoded natural language. We evaluate CodeFusion on the task of natural language to code generation for Bash, Python, and Microsoft Excel conditional formatting (CF) rules. Experiments show that CodeFusion (75M parameters) performs on par with state-of-the-art auto-regressive systems (350M-175B parameters) in top-1 accuracy and outperforms them in top-3 and top-5 accuracy due to its better balance in diversity versus quality.
Metadata: arXiv. Abstracts retain their original authors’ rights.
In the conversation
Original HN comments, with labels from simple text rules. A label describes the whole comment and may miss context or irony. Read the discussion to make up your own mind. About the labels
To be precise, ChatGPT 3.5 turbo being 20B is officially a mistake from a Microsoft Researcher, quoting a wrong source published before the release of chatgpt3.5 turbo. Up to you to believe it or not. But I wouldn’t claim it’s a 20B according to Microsoft Researchers.
The withdrawn paper: https://arxiv.org/abs/2310.17680
The wrong source: https://www.forbes.com/sites/forbestechcouncil/2023/02/17/is...
The discussion: https://www.reddit.com/r/LocalLLaMA/comments/17jrj82/new_mic...
The paper was https://arxiv.org/abs/2310.17680
It has been withdrawn with this note:
> Contains inappropriately sourced conjecture of OpenAI's ChatGPT parameter count from this http URL, a citation which was omitted. The authors do not have direct knowledge or verification of this information, and relied solely on this article, which may lead to public confusion
(the noted URL is a just a Forbes blogger with no special qualifications that would make what he claimed particularly credible).
Link to the paper directly https://arxiv.org/pdf/2310.17680.pdf
The system that they describe, Codefusion is interesting because it's a diffusion model for generating code rather than an autoregressive model like most LLM code generators.
Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task.
Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance.
In honor of Alan Turing's hundredth birthday, I unwisely set out some thoughts about one of Turing's obsessions throughout his life, the question of physics and free will.