Summary of Skillaggregation: Reference-free Llm-dependent Aggregation, by Guangzhi Sun et al.
SkillAggregation: Reference-free LLM-Dependent Aggregation
by Guangzhi Sun, Anmol Kagrecha, Potsawee Manakul, Phil Woodland, Mark Gales
First submitted to arxiv on: 14 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG)
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, SkillAggregation, leverages predictions from multiple Large Language Models (LLMs) to aggregate judgments without requiring additional data or ground truth. Building upon the Crowdlayer aggregation method, which was developed for image classification, this approach extends its capabilities to exploit judge estimates during inference. The performance of SkillAggregation is compared to various standard aggregation methods on three tasks: HaluEval-Dialogue, TruthfulQA, and Chatbot Arena. Results show that SkillAggregation outperforms Crowdlayer on all tasks and achieves the best performance among all approaches for most tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary SkillAggregation is a new way to combine predictions from multiple large language models (LLMs) without needing extra information or correct answers. This helps improve how well LLMs work together to make decisions. The method is tested on three different types of tasks and performs better than other methods for most of them. |
Keywords
* Artificial intelligence * Image classification * Inference
