Summary of Which Llm to Play? Convergence-aware Online Model Selection with Time-increasing Bandits, by Yu Xia et al.
Which LLM to Play? Convergence-Aware Online Model Selection with Time-Increasing Bandits
by Yu Xia, Fang Kong, Tong Yu, Liya Guo, Ryan A. Rossi, Sungchul Kim, Shuai Li
First submitted to arxiv on: 11 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Machine Learning (stat.ML)
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 proposed method addresses the challenge of online model selection for large language models (LLMs) in web-based applications. With the rise of LLMs, organizations face decisions on whether to use costly API-based LLMs or locally finetuned small LLMs, balancing task reward and exploration cost. Traditional methods evaluate every candidate model before choosing one, which is impractical given the increasing costs of training and finetuning LLMs. The proposed method leverages online bandit algorithms to manage the exploration-exploitation trade-off in model selection, taking into account the increasing-then-converging trend in model performances as the model is iteratively finetuned. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Online model selection for large language models (LLMs) helps organizations choose the best LLM for their task. This is important because LLMs are used in many web-based applications like chatbots and search engines. The problem is that choosing the right LLM can be expensive and time-consuming. Some methods try to solve this problem by testing every possible LLM, but this isn’t practical anymore since training and finetuning LLMs takes a lot of resources. |
