Reproduce “Stephanie2: Thinking, Waiting, and Making Decisions Like Humans in Step-by-Step AI Social Chat” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hao Yang, Hongyuan Lu, Dingkang Yang, Wenliang Yang, Peng Sun, Xiaochuan Zhang, Jun Xiao, Kefan He and 3 more
Stephanie2 introduces a dialogue agent designed to enhance human-like social chat by incorporating active waiting and adaptive message pacing in step-by-step interactions. It achieved a 49.60% to 56.24% pass rate in a Turing-test variant, making its dialogues harder to distinguish from human conversations, and reduced average words per message (7.29) closer to human norms (5.84).
Key insights
- Explicitly modeling active waiting and adaptive message pacing significantly improves the naturalness, engagement, and human-likeness of AI social chat interactions.
- Allowing AI to perform a guided "thinking trace" and make explicit `send`/`wait` decisions based on dialogue context enables more socially appropriate and less interruptive conversational behavior.
- Utilizing a "time-window" mechanism in dual-agent dialogue generation fosters more realistic, multi-message conversational bursts, preventing unnatural ping-pong turn-taking.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.