Summary of Large Language Models For Automatic Milestone Detection in Group Discussions, by Zhuoxu Duan et al.
Large Language Models for Automatic Milestone Detection in Group Discussions
by Zhuoxu Duan, Zhengye Yang, Samuel Westby, Christoph Riedl, Brooke Foucault Welles, Richard J. Radke
First submitted to arxiv on: 16 Jun 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
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 explores the performance of large language models (LLMs) on recordings of group oral communication tasks. The experiment involves a puzzle with milestones that can be achieved in any order, making it challenging for LLMs to track progress. To analyze this, the authors propose methods for processing transcripts to detect milestone completion and investigate two approaches: iteratively prompting GPT with transcription chunks and semantic similarity search using text embeddings. Results show that iterative prompting outperforms semantic similarity search, highlighting the importance of context in understanding language models’ responses. The study discusses the quality and randomness of GPT responses under different context window sizes. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how well big language computers do on a new kind of challenge: understanding conversations between groups of people. They create a puzzle that group members can solve in any order, making it hard for computers to keep track of progress. To see how language models work, the researchers propose ways to analyze transcripts and figure out when someone completes a milestone. They test two approaches: asking GPT questions about parts of the conversation and comparing text similarities. The results show that asking GPT questions works better than comparing texts. This study helps us understand how language computers respond differently depending on the context. |
Keywords
» Artificial intelligence » Context window » Gpt » Prompting
