Flexible tactile sensor has emerged as a key component in wearable electronics, intelligent robotics et al., owing to the excellent mechanical compliance and multimodal sensing capabilities. Although the application for solid detection is demonstrated, high-accurate recognition of liquid properties remains a critical challenge due to their inherent fluidity and complex thermophysical characteristics. Herein, a new-style dual-mode passive flexible tactile sensor is presented via integrating an Ag2Se@PU thermoelectric foam with a thermochromic TCPs substrate and a TPU protective layer, capable of simultaneously capturing electrical and visual signals associated with the temperature variations in liquids under a non-contact mode, showing a minimal temperature resolution of 0.3 K. In addition, this device exhibits excellent flexibility, hydrophobicity and mechanical durability, especially maintains high thermoelectric response stability after over 5000 cycles. In practical liquid classification, this sensor can accurately distinguish different drinking water and methanol-ethanol mixtures, achieving ultra-high recognition accuracy of 100% by employing a convolutional neural network trained on the fused thermoelectric and colorimetric dual-mode signals, which outperforms single-mode approaches in identifying highly similar liquid compositions. This work provides a feasible strategy, by combining material-level optimization and multimodal data integration, for high-precision liquid analysis in wearable diagnostics, intelligence robots and so on.

URL：https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202514339