Abstract: Systems and methods are provided for efficiently calibrating a data mixture for training machine-learned models (e.g., machine-learned sequence processing models, such as transformer-based models). For example, machine-learned models can be trained over a broad dataset that can include multiple different categories of data. The mixture of data categories within the dataset can influence model performance. To improve the performance of machine-learned models, example implementations of the present disclosure can learn a distribution of data categories using a lightweight proxy model before initiating training of a large primary model. In this manner, for instance, example implementations can obtain an improved training data distribution with less computational expense and can leverage the learned training data distribution to better train a large primary model.

Abstract: A method for forming a semiconductor device and a semiconductor device formed by the method are disclosed. In an embodiment, the method includes depositing a dummy dielectric layer on a fin extending from a substrate; depositing a dummy gate seed layer on the dummy dielectric layer; reflowing the dummy gate seed layer; etching the dummy gate seed layer; and selectively depositing a dummy gate material over the dummy gate seed layer, the dummy gate material and the dummy gate seed layer constituting a dummy gate.

Abstract: An apparatus includes a processor core and a memory hierarchy. The memory hierarchy includes main memory and one or more caches between the main memory and the processor core. A plurality of hardware pre-fetchers are coupled to the memory hierarchy and a pre-fetch control circuit is coupled to the plurality of hardware pre-fetchers. The pre-fetch control circuit is configured to compare changes in one or more cache performance metrics over two or more sampling intervals and control operation of the plurality of hardware pre-fetchers in response to a change in one or more performance metrics between at least a first sampling interval and a second sampling interval.

Abstract: A method includes placing an electronic die and a photonic die over a carrier, with a back surface of the electronic die and a front surface of the photonic die facing the carrier. The method further includes encapsulating the electronic die and the photonic die in an encapsulant, planarizing the encapsulant until an electrical connector of the electronic die and a conductive feature of the photonic die are revealed, and forming redistribution lines over the encapsulant. The redistribution lines electrically connect the electronic die to the photonic die. An optical coupler is attached to the photonic die. An optical fiber attached to the optical coupler is configured to optically couple to the photonic die.

Abstract: A wireless tire pressure monitoring system and a method for operating the tire pressure monitoring system to locate each wireless tire pressure sensor. The wireless host is connected to the wireless receiver. Each wireless tire pressure sensor sends info to the wireless host and the wireless receiver. The wireless host and the wireless receiver allocate the positions of the wheels where the wireless tire pressure sensors are installed by the info received from wireless tire pressure sensors. The wireless receiver analyses and calculates the received info, and forwards the result to the wireless host to process a cross-comparison to precise allocate the positions of the wireless tire pressure sensors. The method does not need to replace or add new wireless tire pressure sensors, and does not need to obtain new serial numbers. The positions of the wireless tire pressure sensor are precisely allocated by the method of the present invention.

Abstract: An electromagnetic wave guidance and beam reshaping structure is favorable to incorporate a radiation source antenna into an energy focusing system. The electromagnetic wave guidance and beam reshaping structure includes a substrate, a plurality of metal patterns and a plurality of hollow structures. The substrate includes a central portion and a peripheral portion that surrounds the central portion. The plurality of metal patterns are disposed on the central portion. The plurality of hollow structures are disposed in the peripheral portion. The metal patterns are axisymmetrically arranged with respect to a central axis of the substrate, and the hollow structures are axisymmetrically arranged with respect to the central axis of the substrate.

Abstract: Some embodiments of the present disclosure provide a deep-learning-based framework for designing components of a downlink precoding system. The components of such a system include downlink training pilots and channel estimation based on receipt of the downlink training pilots. Another component involves channel measurement and feedback strategy at the user equipment. The components include a precoding scheme designed at the base station based on the feedback from the user equipment.

Abstract: A weft-knitted fabric with cut-loop piles comprises a plurality of basic yarn loops and a plurality of cut-loop piles. The plurality of basic yarn loops are formed from at least one basic yarn, the plurality of cut-loop piles are respectively connected to one of the plurality of basic yarn loops, each of the plurality of cut-loop piles is formed by cutting a pile loop formed from one of at least two colored yarns, each of the at least two colored yarns without forming the plurality of cut-loop piles forms at least one of the plurality of basic yarn loops together with the at least one basic yarn, and colors of the at least two colored yarns are different.

Abstract: A method for handling a display control of a microprocessor in an electronic device includes: receiving a display trigger signal; and controlling a panel device in the electronic device to display a content, in response to the display trigger signal; wherein a central processing unit (CPU) in the electronic device is in a power off state, when controlling the panel device to display the content.

Abstract: Provided is an electronic package, in which a conductive structure and an encapsulation layer covering the conductive structure are arranged on one side of a carrier structure having a circuit layer, and an electronic component is arranged on the other side of the carrier structure. The rigidity of the carrier structure is increased by the encapsulation layer, and problems such as warpage or wavy deformations caused by increasing the volume of the electronic package due to functional requirements can be eliminated.

Abstract: A biomimetic turtle includes a trunk, a head movably connected with a front end of the trunk in the front-rear direction, two front limbs disposed on a front section of the trunk and each rotatable relative to the trunk to sway in an up-down direction, and a driving module to drive the head and the front limbs. Each front limb has a curve-shaped rigid portion with a recess, and a deformable flipper portion engaged in the recess and extending rearwardly. With the deformable flipper portion deformed and bent during swaying of the front limbs, a forward propelling force is generated to propel the biomimetic turtle forwardly. The head is operably movable to vary the center of gravity of 10 the biomimetic turtle 100 in the water, and thus the front portion of the biomimetic turtle is inclined upwardly or downwardly to facilitate ascending or descending of the biomimetic turtle.

Abstract: An image processing system includes: a first image processing device for performing a first image enhancement process on a source image to generate a first enhanced image; one or more second images processing device, each of which is used to perform a second image enhancement processing on a size-reduced image generated based on the source image, and accordingly to output one or more second enhanced images whose size identical to the source image; and an output controller for analyzing regional frequency characteristics of the source image to generate an analysis result, determining one or more region weights according to the analysis result, and synthesize the first enhanced image with the one or more second enhanced images according to the one or more region weights, thereby to generate an output image.

Abstract: Implementations relate to managing multimedia content that is obtained by large language model(s) (LLM(s)) and/or generated by other generative model(s). Processor(s) of a system can: receive natural language (NL) based input that requests multimedia content, generate a response that is responsive to the NL based input, and cause the response to be rendered. In some implementations, and in generating the response, the processor(s) can process, using a LLM, LLM input to generate LLM output, and determine, based on the LLM output, at least multimedia content to be included in the response. Further, the processor(s) can evaluate the multimedia content to determine whether it should be included in the response. In response to determining that the multimedia content should not be included in the response, the processor(s) can cause the response, including alternative multimedia content or other textual content, to be rendered.

Abstract: A processing apparatus includes a processing resource including a general-purpose parallel processing engine and a matrix accelerator. The matrix accelerator includes first circuitry to receive a command to perform operations associated with an instruction, second circuitry to configure the matrix accelerator according to a physical depth of a systolic array within the matrix accelerator and a logical depth associated with the instruction, third circuitry to read operands for the instruction from a register file associated with the systolic array, fourth circuitry to perform operations for the instruction via one or more passes through one or more physical pipeline stages of the systolic array based on a configuration performed by the second circuitry, and fifth circuitry to write output of the operations to the register file associated with the systolic array.

Abstract: An optical photographing lens assembly includes, in order from an object side to an image side, a first lens element, a second lens element, a third lens element, a fourth lens element, a fifth lens element and a sixth lens element. The first lens element with positive refractive power has an object-side surface being convex in a paraxial region thereof. The second lens element has refractive power. The third lens element has refractive power. The fourth lens element has refractive power. The fifth lens element with refractive power has an image-side surface being convex in a paraxial region thereof, wherein an object-side surface and the image-side surface of the fifth lens element are both aspheric. The sixth lens element with refractive power has an object-side surface and an image-side surface being both aspheric. The optical photographing lens assembly has a total of six lens elements with refractive power.

Abstract: A semiconductor package includes a semiconductor element, at least one electronic die, at least one optical die, an encapsulant, and a substrate. The semiconductor element has a first side and a second side opposing to the first side. The at least one electronic die is disposed over the first side. The at least one optical die is disposed over the first side and next to the at least one electronic die. The encapsulant is disposed on the first side and covers the at least one electronic die, where a sidewall of the at least one optical die is distant from the encapsulant, and a sidewall of the encapsulant is aligned with a sidewall of the semiconductor element. The substrate is disposed over the second side, where the at least one electronic die is electrically coupled to the substrate and the at least one optical die through the semiconductor element.

Abstract: A method and an apparatus of recognizing a package sorting behavior are provided. A target detection is performed on an image frame in a target video, so as to acquire at least one human detection box and at least one package detection box; a motion trajectory of the human detection box and a motion trajectory of the package detection box are tracked respectively; in a process of trajectory tracking for a package, a moment at which the package is thrown is recognized based on a current motion trajectory tracked for the package; and motion information of the package from the moment at which the package is thrown to a current moment is acquired, and a sorting behavior of the package is recognized based on the motion information.

Abstract: A method of handling a workpiece includes the following steps. A workpiece is placed on a chuck body, wherein the workpiece includes a tape carrier extending beyond a periphery of the chuck body and a workpiece body disposed on the tape carrier, and the chuck body includes a seal ring surrounding the periphery of the chuck body; the tape carrier is clamped outside the chuck body, wherein the tape carrier leans against the seal ring and an enclosed space is formed between the chuck body, the tape carrier and the seal ring; and a vacuum seal is formed by evacuating gas from the enclosed space to pull the periphery of the workpiece toward the chuck body.

Abstract: A compact antenna design is provide with a carrier arm having a first end and a second end; a resonating arm, connected at the second end of the carrier arm and extending perpendicularly from the carrier arm in a first direction; a capacitive tuning arm, having a third end and a fourth end, connected at the third end to a portion of the carrier arm between to the first end and the second end, and extending perpendicularly from the carrier arm in the first direction; and a shorting arm, connected at the fourth end of the capacitive tuning arm and extending perpendicularly from the capacitive tuning arm in a second direction, away from the resonating arm.

Abstract: A circular knitting machine used to weave a double-sided plush fabric. The circular knitting machine comprises a horizontal needle bed, a sinker bed and a vertical needle bed. The horizontal needle bed comprises a plurality of hook needles and a plurality of cutters spaced at intervals between the plurality of hook needles, the sinker bed comprises a plurality of sinkers, and the vertical needle bed comprises a plurality of knitting needles with cutters and a plurality of crochet needles spaced at intervals between the plurality of knitting needles with cutters. Before the knitting needles with cutters to be ready for cutting a first pile loop: a latch needle comprised in the plurality of knitting needle with cutters that is ready for cutting yarn comprises been completely cleared, and the plurality of knitting needle with cutters to be ready for cutting yarn passed through at least one plain stitch cam.