Fine-tuned LLMs predict conversation turn-taking from sensor data in MR group tasks at 96% accuracy and 3.2x better than LSTM baselines for linguistic behaviors, but fail on shared attention and degrade sharply in simulation mode.
To- ward sensor-in-the-loop llm agent: Benchmarks and implications,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.HC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
TeamLLM: Exploring the Capabilities of LLMs for Multimodal Group Interaction Prediction
Fine-tuned LLMs predict conversation turn-taking from sensor data in MR group tasks at 96% accuracy and 3.2x better than LSTM baselines for linguistic behaviors, but fail on shared attention and degrade sharply in simulation mode.