The Future of AI Interaction: True Multimodal Conversations
Date:
This presentation introduces our work on test-time scaling strategies for generative retrieval in multimodal conversational recommendations. We demonstrate how test-time reranking can significantly improve conversational product retrieval with minimal overhead.
Key Topics Covered
- Introduction to test-time scaling in conversational search
- Challenges with existing multimodal retrievers in iterative dialogues
- Test-time reranking (TTR) mechanism design
- Experimental results showing 14.5 points improvement in MRR and 10.6 points in nDCG@1
- Enhanced datasets for future research
The slides provide a comprehensive overview of our methodology and experimental findings in this important but underexplored area of conversational AI.