Recent advancements in Large Language Models (LLMs) have significantly propelled the development of Conversational Recommendation Agents (CRAs). However, these agents often generate short-sighted responses that fail to sustain user guidance and meet expectations. Although preference optimization has proven effective in aligning LLMs with user expectations, it remains costly and performs poorly in multi-turn dialogue. To address this challenge, we introduce a novel multi-turn preference optimization (MTPO) paradigm **ECPO**, which leverages Expectation Confirmation Theory to explicitly model the evolution of user satisfaction throughout multi-turn dialogues, uncovering the underlying causes of dissatisfaction. These causes can be utilized to support targeted optimization of unsatisfactory responses, thereby achieving turn-level preference optimization. ECPO eliminates the significant sampling overhead of existing MTPO methods while ensuring the optimization process drives meaningful improvements. To support ECPO, we also introduce an LLM-based user simulator, **AILO**, to simulate user feedback and expectation confirmation during conversational recommendations. Experimental results show that ECPO significantly enhances CRA’s interaction capabilities, offering notable improvements in both efficiency and effectiveness over existing MTPO methods.

The task shared by sponsor about Hope Speech Detection for Equality, Diversity, and Inclusion at LT-EDI-ACL-2022.The goal of this task is to identify whether a given comment contains hope speech or not,and hope is considered significant for the well-being, recuperation and restoration of human life. Our work aims to change the prevalent way of thinking by moving away from a preoccupation with discrimination, loneliness or the worst things in life to building the confidence, support and good qualities based on comments by individuals. In response to the need to detect equality, diversity and inclusion of hope speech in a multilingual environment, we built an integration model and achieved well performance on multiple datasets presented by the sponsor and the specific results can be referred to the experimental results section.