Large Language Models Empower Personalized Valuation in Auction

Jie Sun,Tianyu Zhang,Houcheng Jiang,Kexin Huang,Chi Luo,Junkang Wu,Jiancan Wu,An Zhang,Xiang Wang

2024-10-21

Abstract:Auctions, a fundamental economic mechanism, encompass the valuation of goods or services and the competitive bidding algorithms within a specific framework, serving to uncover the true market value. However, current research predominantly focuses on the bidding algorithms within a given auction mechanism, often overlooking the advantages of incorporating individual bidders' unique preferences and the semantic information related to the items into the valuation process. Our analysis, both theoretical and empirical, shows that imprecise or noisy valuations can significantly affect the overall utility for participants. To bridge this gap, we propose a personalized valuation framework, namely \textbf{S}emantic-enhanced \textbf{P}ersonalized \textbf{V}aluation in \textbf{A}uction (\ours), which integrates Large Language Models (LLMs) to incorporate semantic information into each bidder's unique valuation process. Specifically, SPVA employs a two-stage approach: it first fine-tunes LLMs to encode bidder preferences in personalized valuations, and then constructs a Vickrey auction environment integrated with a bidding algorithm to demonstrate that SPVA's more accurate valuations result in higher profits. Additionally, we have developed a semantic-enhanced dataset comprising over 23,000 samples and introduced new personalized evaluation metrics that reflect both bidder preferences and profit. Through simulations of various auction scenarios, our method demonstrates its ability to provide accurate valuations and capture bidder preferences, affirming the method's effectiveness in real-world auction settings.

Computational Engineering, Finance, and Science

What problem does this paper attempt to address?

This paper attempts to solve the problem of personalized valuation in auction mechanisms. Specifically, existing auction research mainly focuses on bidding algorithms within a given auction mechanism, often ignoring the incorporation of bidders' unique preferences and item - related semantic information into the valuation process. This leads to inaccurate or noisy valuations, which significantly affect the overall utility of participants.

To address these issues, the authors propose a framework named **Semantic - enhanced Personalized Valuation in Auction (SPVA)**, which captures bidders' unique preferences and item semantics by integrating large - language models (LLMs). SPVA adopts a two - stage approach: first, fine - tune LLMs to encode bidders' personalized preferences; then, combine with bidding algorithms in the Vickrey auction environment to show that more accurate valuations can lead to higher profits.

### Main Contributions
1. **Developed a unique dataset specifically for auction valuation**, which covers detailed item characteristics and market values.
2. **Proposed and proved a theorem**, indicating that when valuations are affected by noise, existing bidding algorithms lead to a decline in utility.
3. **Proposed the SPVA framework**, which utilizes the capabilities of large - language models for detailed data interpretation to achieve semantically enhanced personalized valuation.
4. **Introduced new evaluation metrics, Essential Utility (EU) and Essential Value (EV)**, for accurately assessing the accuracy of bidders' personalized valuations and the accuracy of their decisions.

### Method Overview
- **Data Pre - processing**: Solved three main problems in the original Epinions dataset through GPT - 4: the difference between price and valuation, rating bias, and lack of explanatory information.
- **Personalized Valuation**: Used GPT - 4 to generate diverse instructions and reasons, and fine - tuned the Mistral model through LoRA technology to enable it to handle valuation commands in auctions.
- **Auction Bidding**: Simulated the Vickrey auction mechanism, combined with the Individual Pacing algorithm, to help bidders formulate bidding strategies and calculate auction results.

### Experimental Verification
Through noise experiments under different standard deviations, the impact of noise on utility was proven. The experimental results show that as the standard deviation of noise increases, the utility decreases significantly. In addition, the SPVA framework demonstrated its ability to increase profits in various auction scenarios.

In conclusion, this paper aims to improve the accuracy and effectiveness of personalized valuation in auction mechanisms by introducing large - language models and semantic information, thereby optimizing bidders' decisions and overall returns.