Abstract: A method for LLM inference is provided. The method includes: receiving an input prompt; generating sequentially multiple draft tokens based on the input prompt by a draft model; generating multiple answer tokens in parallel based on the multiple draft tokens by a target model; and verifying at least one of the draft tokens. The verification result indicates that, when a ratio of a first probability of one of the draft tokens to a second probability of the one of the draft tokens is larger than a threshold which is lower than 1, or the one of the draft tokens is identical to a corresponding one of the answer tokens, the one of the draft tokens passes the verification.

Abstract: An easy-disconnect seal matching reservoir for connecting the microfluidic chip and the pipette. The reservoir comprises at least one first coupling unit and at least one second coupling unit. The first coupling unit includes a first end and a second end opposed to the first end, and the first end is disposed on an inlet of the microfluidic chip. The second coupling unit includes a third end and a fourth end opposed to the third end, and a pipe connected between the third end and the fourth end. The third end is coupled to the second end of the first coupling unit. The pipette can be put in the pipe via the fourth end.

Abstract: An anti-loose interdental brush includes a plurality of fixing portions formed on a bristle stem of an interdental brush. Each fixing portion includes a bottom wall, a side wall formed laterally from the bottom wall, and a capacity area surrounded by the bottom wall and the side wall. The adjacent bottom walls of the adjacent fixing portions are orientated in different directions. Thus, the adjacent capacity areas are set in different directions to contact the handle. An outer diameter of the bottom wall can be preferably larger than or equal to an outer diameter of the bristle stem. Accordingly, a dual engaging and fixing effect between the bristle stem and the handle is attained to prevent the bristle stem from loosening and rotating due to an external force and to enhance safety in using.

Abstract: An anti-loose interdental brush comprises a plurality of fixing portions formed on a bristle stem of an interdental brush. Each fixing portion includes a bottom wall, a side wall formed laterally from the bottom wall, and a capacity area surrounded by the bottom wall and the side wall. The adjacent bottom walls of the adjacent fixing portions are orientated in different directions, whereby the adjacent capacity areas are set in different directions to contact said handle. An outer diameter of the bottom wall can be preferably larger than or equal to an outer diameter of the bristle stem. Accordingly, a dual engaging and fixing effect between the bristle stem and the handle is attained to prevent the bristle stem from loosening and rotating due to an external force and enhance the safety in using.

Abstract: An object detection method and an object detection system, suitable for detecting moving object information of a video stream having a plurality of images, are provided. The method performs a moving object foreground detection on each of the images, so as to obtain a first foreground detection image comprising a plurality of moving objects. The method also performs a texture object foreground detection on each of the images, so as to obtain a second foreground detection image comprising a plurality of texture objects. The moving objects in the first foreground detection image and the texture objects in the second foreground detection image are selected and filtered, and then the remaining moving objects or texture objects after the filtering are output as real moving object information.

Abstract: An object detection method and an object detection system, suitable for detecting moving object information of a video stream having a plurality of images, are provided. The method performs a moving object foreground detection on each of the images, so as to obtain a first foreground detection image comprising a plurality of moving objects. The method also performs a texture object foreground detection on each of the images, so as to obtain a second foreground detection image comprising a plurality of texture objects. The moving objects in the first foreground detection image and the texture objects in the second foreground detection image are selected and filtered, and then the remaining moving objects or texture objects after the filtering are output as real moving object information.

Abstract: A multi-state target tracking method and a multi-state target tracking system are provided. The method detects a crowd density of a plurality of images in a video stream and compares the detected crowd density with a threshold when receiving the video stream, so as to determine a tracking mode used for detecting the targets in the images. When the detected crowd density is less than the threshold, a background model is used to track the targets in the images. When the detected crowd density is greater than or equal to the threshold, a none-background model is used to track the targets in the images.