Summary of Assessing the Emergent Symbolic Reasoning Abilities Of Llama Large Language Models, by Flavio Petruzzellis et al.
Assessing the Emergent Symbolic Reasoning Abilities of Llama Large Language Models
by Flavio Petruzzellis, Alberto Testolin, Alessandro Sperduti
First submitted to arxiv on: 5 Jun 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
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 The paper investigates the capabilities and limitations of popular open-source Large Language Models (LLMs) on symbolic reasoning tasks. Specifically, it evaluates three LLMs from the Llama 2 family – a generalist model (Llama 2 Chat) and two fine-tuned models (MAmmoTH and MetaMath) designed for mathematical problem-solving. The study uses two datasets featuring mathematical formulas of varying difficulty to test the models’ performance. Results show that increasing model scale and fine-tuning on relevant tasks lead to significant gains, with larger models performing better on simpler math problems. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper looks at how big language models do on math problems. It tests three different types of these models: one that’s good at general things, and two that were trained specifically for math. The models are given math problems of different difficulty to solve. The results show that the bigger models and those trained for math get better at solving simple math problems. |
Keywords
» Artificial intelligence » Fine tuning » Llama
