Abstract: A method includes forming a reconstructed wafer, which includes placing a plurality of device dies over a carrier, encapsulating the plurality of device dies in an encapsulant, and forming a redistribution structure over the plurality of device dies and the encapsulant. The redistribution structure includes a plurality of dielectric layers and a plurality of redistribution lines in the plurality of dielectric layers. The method further includes performing a trimming process on the reconstructed wafer. The trimming process forms a round edge for the reconstructed wafer. A sawing process is performed on the reconstructed wafer, so that the reconstructed wafer includes straight edges.

Abstract: A method includes obtaining, using a first electronic device, a weight matrix associated with a trained transformer model. The method also includes factorizing the weight matrix into a dictionary weight matrix and an intermediate matrix. The method further includes pruning the intermediate matrix to generate a sparse intermediate matrix. The method also includes fine-tuning the sparse intermediate matrix based on a training dataset to generate a fine-tuned sparse intermediate matrix. The method further includes determining an index matrix and a coefficient matrix based on the fine-tuned sparse intermediate matrix. In addition, the method includes deploying the dictionary weight matrix, the index matrix, and the coefficient matrix to a second electronic device without deploying the weight matrix to the second electronic device. A number of parameters in the dictionary weight matrix, the index matrix, and the coefficient matrix is smaller than a number of parameters in the weight matrix.

Abstract: A poly(amide-imide) is provided. The poly(amide-imide) is represented by formula (1), wherein R is a C6 aryl group, a C7-C8 aralkyl group, a C2-C6 alkoxyalkyl group, or a C3-C18 alkyl group; and 0.02?X?0.5.

Abstract: An isolation structure can be formed between adjacent and/or non-adjacent pixel regions (e.g., between diagonal or cross-road pixel regions), of an image sensor, to reduce and/or prevent optical crosstalk. The isolation structure may include a deep trench isolation (DTI) structure or another type of trench that is partially filled with a material such that an air gap is formed therein. The DTI structure having the air gap formed therein may reduce optical crosstalk between pixel regions. The reduced optical crosstalk may increase spatial resolution of the image sensor, may increase overall sensitivity of the image sensor, may decrease color mixing between pixel regions of the image sensor, and/or may decrease image noise after color correction of images captured using the image sensor.

Abstract: A manufacturing method of a semiconductor package includes the following steps. At least one lower semiconductor device is provided. A plurality of conductive pillars are formed on the at least one lower semiconductor device. A dummy die is disposed on a side of the at least one lower semiconductor device. An upper semiconductor device is disposed on the at least one lower semiconductor device and the dummy die, wherein the upper semiconductor device reveals a portion of the at least one lower semiconductor device where the plurality of conductive pillars are disposed. The at least one lower semiconductor device, the dummy die, the upper semiconductor device, and the plurality of conductive pillars are encapsulated in an encapsulating material. A redistribution structure is formed over the upper semiconductor device and the plurality of conductive pillars.

Abstract: A chip package structure includes an interposer structure that contains a package-side redistribution structure, an interposer core assembly, and a die-side redistribution structure. The interposer core assembly includes at least one silicon substrate interposer, and each of the at least one silicon substrate interposer includes a respective silicon substrate, a respective set of through-silicon via (TSV) structures vertically extending through the respective silicon substrate, a respective set of interconnect-level dielectric layers embedding a respective set of metal interconnect structures, and a respective set of metal bonding structures that are electrically connected to the die-side redistribution structure. The chip package structure includes at least two semiconductor dies that are attached to the die-side redistribution structure, and an epoxy molding compound (EMC) multi-die frame that laterally encloses the at least two semiconductor dies.

Abstract: Disclosed herein are humanized antibodies, antigen-binding fragments thereof, and antibody conjugates, that are capable of specifically binding to certain biantennary Lewis antigens, which antigens are expressed in a variety of cancers. The presently disclosed antibodies are useful to target antigen-expressing cells for treatment or detection of disease, including various cancers. Also provided are polynucleotides, vectors, and host cells for producing the disclosed antibodies and antigen-binding fragments thereof. Pharmaceutical compositions, methods of treatment and detection, and uses of the antibodies, antigen-binding fragments, antibody conjugates, and compositions are also provided.

Abstract: The fiber optic connector includes a connector head module, a mounting seat, a rear boot, an engaging module and a sheath member. The mounting seat is mounted to a rear end of the connector head module, and includes an external threaded portion. The rear boot is connected to a rear end of the mounting seat. The engaging module is removably coupled to the connector head module. The sheath member includes an internal threaded portion that is formed in an inner surface of the sheath member. When the engaging module is removed from the connector head module, the sheath member can be attachable to the mounting seat with the external threaded portion being threadedly engaged with the external threaded portion of the mounting seat.

Abstract: A use of extracts of roselle seeds for preparing a composition of antioxidant, anti-inflammation, anti-UVB, anti-allergy, whitening, moisturizing, and anti-wrinkle, wherein the extracts of the roselle seeds comprise lauric, palmitic, linoleic, oleic, or stearic acids; and a method for preparing the extracts of the roselle seeds comprises steps of: steeping of: extracting a weight ratio of the 12:1 to 20:1 roselle seeds and an ethanol solution with a concentration of 75 to 95%; and steeping the roselle seeds in the ethanol solution to obtain a suspension; and extraction of: the suspension being engaged in an extraction of a continuous and rapid oscillation with an ultrasonic energy with a total energy of 200 to 500W, wherein 1 to 3 hours of the oscillation is one cycle, and a total number of 3 to 5 cycles is lasted, in order to obtain roselle seed crude extracts.

Abstract: A method of preparing a thermoplastic composition is provided. The method includes the following steps. A polyetherimide or a polyphenylene sulfide is provided. A polyimide is provided, wherein the glass transition temperature of the polyimide is between 128° C. and 169° C., the 10% thermogravimetric loss temperature of the polyimide is between 490° C. and 534° C., and when the polyimide is dissolved in N-methyl-2-pyrrolidone and the solid content of the polyimide is 30 wt %, the viscosity of the polyimide is between 100 cP and 250 cP. A melt process is performed to mix the polyetherimide and the polyimide or mix the polyphenylene sulfide and the polyimide to form a thermoplastic composition. Further, a thermoplastic composition is also provided.

Abstract: A power adapter assembly structure is disclosed and includes a circuit board, a socket and at least one elastic element. The socket is disposed adjacent to the circuit board. The circuit board and the socket are configured to collaboratively form at least one abutting surface and at least one fixing surface. The elastic element is connected between the circuit board and the socket, and includes a main body, a fixed portion and a hanging arm. The fixed portion and the hanging arm are disposed at two opposite ends of the main body, the fixed portion spatially corresponds to the fixing surface, and the hanging arm constantly abuts the abutting surface. A height is formed between the main body of the at least one elastic element and the at least one abutting surface, and less than a length of the hanging arm extended from the main body.

Abstract: A power adapter assembly structure is disclosed and includes a housing, a circuit board, a socket and an elastic element. The circuit board including a first abutting surface and the socket including a second abutting surface are fastened in the housing. The elastic element connected between the circuit board and the socket includes a main body, a first hanging arm and a second hanging arm. The first hanging arm and the second hanging arm are disposed at two opposite ends of the main body, and constantly abuts the first abutting surface and the second abutting surface, respectively. A first gap is formed between the main body and the first abutting surface, and less than a first length of the first hanging arm. A second gap is formed between the main body and the second abutting surface, and less than a second length of the second hanging arm.

Abstract: A method includes receiving one or more training corpora for training a machine learning model having a plurality of encoder blocks, where each encoder block includes an attention layer and a feedforward network. The method also includes using the one or more training corpora to train an attention dictionary shared across the plurality of encoder blocks. Training the attention dictionary may include training attention parameters of the attention layer in each of the plurality of encoder blocks, and the attention parameters for a given encoder block among the plurality of encoder blocks may be a weighted combination of columns from the attention dictionary shared across the plurality of encoder blocks.

Abstract: A method includes obtaining a parameter matrix associated with a linear layer of a first machine learning model and containing parameter values for parameters of the linear layer. The method also includes determining importance values corresponding to the parameter values. The method further includes generating factorized matrices such that a product of the importance values and factorized matrices contains approximated parameter values for the parameters of the linear layer. In addition, the method includes generating a second machine learning model representing a compressed version of the first machine learning model. The second machine learning model has first and second linear layers containing parameter values based on the importance values and the factorized matrices. The factorized matrices are generated based on weighted errors between the parameter values for the parameters of the linear layer and the approximated parameter values.

Abstract: A method includes providing, using at least one processing device of an electronic device, input data to a machine learning model. The method also includes extracting, using the at least one processing device, features of the input data. The method further includes performing, using the at least one processing device, a geometric transformation of the features, where the geometric transformation is based on first and second parametric instance-dependent scalar functions. In addition, the method includes producing, using the at least one processing device, a predictive probability distribution based on the transformed features.

Abstract: There is provided a smoke detector including a light source, a reflective surface, a light sensor and a processor. The light sensor receives reflected light when the light source emits light toward the reflective surface, and generates a reference detection signal when there is no smoke. The processor receives the detection signal from the light sensor, and automatically selects a set of predetermined condition thresholds according to a profile of the detection signal to be compared with the detection signal thereby determining whether to generate an alarm according to the comparison result.

Abstract: A USB Type-C male connector includes a first terminal assembly including a plurality of first pins and a second terminal assembly including a plurality of second pins. The first pins of the first terminal assembly having an SMD structure are electrically connected to a substrate in the SMT process. The second pins of the second terminal assembly having a DIP structure are electrically connected to the substrate in the SMT process. More pins of the second terminal assembly are used for transmitting data signals, control signals and power signals, thereby improving the signal quality of the USB drive.

Abstract: A photo-sensitive device includes a uniform layer, a gradated buffer layer over the uniform layer, a silicon layer over the gradated buffer layer, a photo-sensitive light-sensing region in the uniform layer and the silicon layer, a device layer on the silicon layer, and a carrier wafer bonded to the device layer.

Abstract: An interactive device for animals is provided that includes a main body, a driving module and a first rotating member. The main body includes an accommodating groove, an opening, and a communicating channel. The driving module is disposed on the main body. The first rotating member is rotatably disposed in the main body and separates the communicating channel and the accommodating groove. When the driving module drives the first rotating member to rotate in a first rotating direction, the first rotating member drives at least one object disposed in the accommodating groove to enter the communicating channel and leave the main body through the opening.

Abstract: A method for forming a semiconductor device is provided. In some embodiments, the method includes forming a target layer over a semiconductor substrate, forming a carbon-rich hard masking layer over the target layer, patterning features in the carbon-rich hard masking layer using an etching process, performing a directional ion beam trimming process on the features patterned in the carbon-rich hard masking layer, and patterning the target layer using the carbon-rich hard masking layer as a mask.