Summary of Learning to Reason Via Program Generation, Emulation, and Search, by Nathaniel Weir et al.
Learning to Reason via Program Generation, Emulation, and Search
by Nathaniel Weir, Muhammad Khalifa, Linlu Qiu, Orion Weller, Peter Clark
First submitted to arxiv on: 25 May 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper extends the capabilities of language models (LMs) in program synthesis, enabling them to tackle complex tasks such as commonsense reasoning, moral decision-making, and sarcasm understanding. To achieve this, the authors propose Code Generation and Emulated EXecution (CoGEX), a novel approach that trains LMs to generate pseudo-programs, emulate their execution, and search for optimal programs. CoGEX is shown to yield significant improvements over standard in-context learning approaches on both algorithmic and soft reasoning tasks. The authors also release a dataset, fine-tuned models, and implementation for reproducing the results. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps computers learn new skills by writing code that can solve complex problems. Right now, computers are good at solving simple math problems, but they struggle with more human-like tasks like understanding sarcasm or making moral decisions. The authors develop a new way to teach computers to write code for these kinds of tasks using language models. They show that this approach works better than other methods and release the tools they used so others can try it too. |
