Sensor-Language-Action Models
The paper proposes one model that ties sensor readings to language-described actions, rather than stopping at perception.
SLA uses language as the bridge between multimodal sensor observations and action prediction or explanation. The authors report a benchmark covering more than 116,000 individuals, 79 sensor modalities, and 60 action groups. Their OpenSLA model is tested on clinical prediction, operating-room, and metabolic-health tasks, where they say it outperforms prior state of the art. They also report evidence-grounding and zero-shot generalization to unseen actions and cohorts. HF Daily Papers' note
SLA uses language as the bridge between multimodal sensor observations and action prediction or explanation. The authors report a benchmark covering more than 116,000 individuals, 79 sensor modalities, and 60 action groups. Their OpenSLA model is tested on clinical prediction, operating-room, and metabolic-health tasks, where they say it outperforms prior state of the art. They also report evidence-grounding and zero-shot generalization to unseen actions and cohorts. HF Daily Papers' note
score 5