Research map / Large language models / AI and society
Human or Machine? A Preliminary Turing Test for Speech-to-Speech Interaction
Xiang Li, Jiabao Gao, Sipei Lin, Xuan Zhou, Chi Zhang, Bo Cheng, Jiale Han, Benyou Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fine-tune Qwen2.5-Omni with ODL fine-grained scoring and linear classifier instead of off-the-shelf use | Qwen2.5-Omni judge · Turing test judging overall | accuracy: 0.4163 -> 0.9605 | Improved |
| use ODL interpretable judge instead of plain LoRA fine-tuning | Qwen2.5-Omni judge · Turing test judging overall | accuracy: 0.5744 -> 0.9605 | Improved |
| use ODL interpretable judge instead of plain LoRA fine-tuning | Qwen2.5-Omni judge · Pseudo Human dialogues | accuracy: 0.0972 -> 0.9306 | Improved |
| LoRA fine-tuning of Qwen2.5-Omni | Qwen2.5-Omni · Human-Machine dialogues | accuracy: 0.2361 -> 0.6319 | Improved |
| LoRA fine-tuning of Qwen2.5-Omni | Qwen2.5-Omni · Pseudo Human dialogues | accuracy: 0.2361 -> 0.0972 | Hurt |
| use interpretable AI judge instead of human judges | Turing test judge · Overall dialogues | accuracy: 0.7284 -> 0.9605 | Improved |
| evaluate interpretable judge on out-of-domain data | Interpretable AI judge · Overall accuracy inner vs OOD | accuracy: 0.9605 -> 0.9740 | No clear effect |
| higher AI familiarity of participants | human judges · Turing test participant accuracy | accuracy: 64.2% -> 78.8% | Improved |
| increase dialogue length | human judges · Turing test accuracy by duration | accuracy: H-H Z=1.6604, p=0.09683 | No clear effect |
| use text-to-speech pseudo-human dialogues instead of S2S systems | Turing test · Success rate | success rate: 0.07-0.31 for S2S systems | Improved |
Closest papers by meaning
- RealityTest: How People Probe AI Identity and Whether Models Disclose It (NeurIPS 2026 · similarity 0.86)
- Talking Turns: Benchmarking Audio Foundation Models on Turn-Taking Dynamics (ICLR 2025 · similarity 0.86)
- $\tau$-Voice: Benchmarking Full-Duplex Voice Agents on Real-World Domains (ICML 2026 · similarity 0.85)
- A Statistical Framework for Ranking LLM-based Chatbots (ICLR 2025 · similarity 0.85)
- Human Simulacra: Benchmarking the Personification of Large Language Models (ICLR 2025 · similarity 0.85)
- Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond) (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
