This paper addresses the challenge of evaluating the bargaining abilities of Large Language Models (LLMs) in a negotiation context, which is crucial for developing autonomous AI agents. The authors formally define the Bargaining task as an asymmetric incomplete information game, focusing on the gains of the Buyer and Seller in multiple bargaining processes. They collect a real product price dataset, *AmazonHistoryPrice*, and create a benchmark to test various LLMs' performance as Buyers or Sellers. The results show that playing as a Buyer is significantly more difficult than playing as a Seller, and increasing model size does not effectively improve Buyer performance. To address this, the authors propose OG-Narrator, a novel approach that integrates a deterministic Offer Generator and an LLM Narrator to enhance the Buyer's performance. Experimental results demonstrate that OG-Narrator significantly improves the Buyer's deal rates and profits, even for models that have not been aligned. The paper also discusses the limitations and future directions, emphasizing the need for more flexible and effective methods to enhance agents' logic, comprehension, and strategy-making.This paper addresses the challenge of evaluating the bargaining abilities of Large Language Models (LLMs) in a negotiation context, which is crucial for developing autonomous AI agents. The authors formally define the Bargaining task as an asymmetric incomplete information game, focusing on the gains of the Buyer and Seller in multiple bargaining processes. They collect a real product price dataset, *AmazonHistoryPrice*, and create a benchmark to test various LLMs' performance as Buyers or Sellers. The results show that playing as a Buyer is significantly more difficult than playing as a Seller, and increasing model size does not effectively improve Buyer performance. To address this, the authors propose OG-Narrator, a novel approach that integrates a deterministic Offer Generator and an LLM Narrator to enhance the Buyer's performance. Experimental results demonstrate that OG-Narrator significantly improves the Buyer's deal rates and profits, even for models that have not been aligned. The paper also discusses the limitations and future directions, emphasizing the need for more flexible and effective methods to enhance agents' logic, comprehension, and strategy-making.