Welcome to Chirpy Cardinal.
We are a Stanford NLP research group focussed on open-domain conversational dialogue.Our research is informed by hundreds of thousands of people chatting with our bot Chirpy Cardinal via the Alexa Prize, a university competition to advance conversational AI. In 2023, we won 1st Place (Science Prize).
In collaboration with Open Virtual Assistant Lab (OVAL), we have open-sourced our award-winning socialbot from Alexa Prize 2020/2021.
Publications and Media
- SIGDIAL 2022: New research paper on predicting lead-time to initiation.
- ICLR 2022: New research paper on jointly training neural retriever and generator for knowledge-grounded conversations.
- NAACL 2021 and SIGDIAL 2021: 4 new research papers accepted!
- Stanford AI Lab Blog: Inside Chirpy Cardinal: Stanford’s Open-Source Social Chatbot that Won 2nd place in the Alexa Prize
- Stanford HAI: How to Create Better Chatbot Conversations
- Forbes: Talking Machines: Who Won This Year’s Alexa Prize?
- Amazon.science: “This technology will be transformative in ways we can barely comprehend” - includes interviews with faculty advisor Chris Manning, team leads Ashwin Paranjape and Abigail See, and other participants.
- Let’s Talk AI Podcast: Abi and Ashwin discuss their experience of the Alexa Prize, lessons learned, and future paths.
Alexa Prize 5 (2022-2023)
In September 2023, our revamped Chirpy Cardinal bot won 1st Place (Science Prize) in the finals of the Alexa Prize Socialbot Grand Challenge 5. We introduced a number of innovations, including:
- CAMEL: a markup language for stateful, editable, and neurally interpolable chatbot development
- Goldfinch: a method for automatically generates nodes in a conversational graph using an LLM
- Kingfisher: a method for reducing LLM latency by predicting future user utterances and precomputing responses to them
Alexa Prize 4 (2020-2021)
In July 2021, our much improved Chirpy Cardinal again won 2nd place in the finals of the Alexa Prize Socialbot Grand Challenge 4.
The Chirpy Cardinal team worked hard building on the success of the first team. Updates included:
- Working on methods for controllable neural generation
- Updating the previous news, music and movies modules with new APIs, as well as major improvements in entity linking, topical transitions, and latency
- Expanding the variety of responses via new modules that focus on personal issues, sports, food, and even extraterrestrial life!
Alexa Prize 3 (2019-2020)
In July 2020, Chirpy Cardinal won 2nd place (of 10 teams) in the finals of the Alexa Prize Grand Challenge 3.
To find out more about our bot, read our system paper, Neural Generation Meets Real People: Towards Emotionally Engaging Mixed-Initiative Conversations.
Alternatively, you can watch our overview presentation: