Abstract: Embodiments described herein provide a method for mitigating toxic content in text generation by a neural network based framework. The method includes the following operations. A text input of a sequence of tokens is received via a communication interface. A first output probability for a next token generating is generated by a first neural network model that is trained to generate tokens belonging to a prioritized category of vocabulary, in response to the text input. A second output probability of the next token is generated by a second neural network model that is trained to generate tokens belonging to an indiscriminate vocabulary, in response to the text input. The next token for a text output based on a combined output probability computed based on a correction item reflective of the first output probability and the second output probability is generated in response to the text input.

Abstract: Embodiments described herein adopts a k nearest neighbor (kNN) mechanism over a model's hidden representations to identify training examples closest to a given test example. Specifically, a training set of sequences and a test sequence are received, each of which is mapped to a respective hidden representation vector using a base model. A set of indices for each sequence index that minimizes a distance between the respective hidden state vector and a test hidden state vector is then determined A weighted k-nearest neighbor probability score can then be computed from the set of indices to generate a probability distribution over labels for the test sequence.

Abstract: Embodiments described herein provide systems and methods for retrieval augmented generation. A neural network based language model may be provided a question as a user input. Based on the user input, semantically diverse queries may be generated for retrieval from diverse data sources. For example, a structured data source (e.g., database or knowledge base) and unstructured data (e.g., text articles) may be used to retrieve information relevant to the user input. The retrieve information may be ranked so that the most relevant information is used by the language model in generating an answer to the question in the user input. A non-retrieval based answer generated by the language model may be utilized in some embodiments in generating the final answer.

Abstract: A display substrate and a display device are provided. The display substrate includes a base substrate and a plurality of subpixels provided on the base substrate. Each subpixel includes a pixel circuit. The plurality of subpixels include first subpixels. The display substrate further includes power lines. Each power line is configured to provide a first supply voltage for the corresponding first subpixels, the power line is provided on a side of the pixel electrode of the first subpixel near the base substrate, and the power line includes a power line main body and a power line protrusion protruded from the power line main body. The power line protrusion is at least partially overlapped with the pixel electrode of the first subpixel in a direction perpendicular to the base substrate. The display substrate can effectively improve the display quality.

Abstract: The present disclosure relates to a mask, a mask device and a mask design method. The mask includes: a pattern area, wherein the pattern area includes a plurality of openings, among which adjacent openings are spaced apart by a first rib, and at least one of the plurality of openings has a non-straight side, two first straight sides intersecting with the non-straight side, and a second straight side opposite to the non-straight side and intersecting with the first straight sides; and a second rib located at an edge of the pattern area, wherein the second rib is provided with a cutout for compensating for stretch deformation of the non-straight side, extending lines of the two first straight sides extending toward the second rib along a first direction intersect with the second rib to obtain a first area, and the cutout is located at least within the first area.

Abstract: A display substrate and a display device are provided. The display substrate includes a base substrate and a plurality of subpixels provided on the base substrate. Each subpixel includes a pixel circuit. The plurality of subpixels include first subpixels. The display substrate further includes power lines. Each power line is configured to provide a first supply voltage for the corresponding first subpixels, the power line is provided on a side of the pixel electrode of the first subpixel near the base substrate, and the power line includes a power line main body and a power line protrusion protruded from the power line main body. The power line protrusion is at least partially overlapped with the pixel electrode of the first subpixel in a direction perpendicular to the base substrate. The display substrate can effectively improve the display quality.

Abstract: A film mask, a manufacturing method thereof, a display panel and a display device are provided. The mask includes a frame and a first align mask. The frame includes two first frame edges extending along a first direction and arranged at intervals along a second direction and two second frame edges extending along the second direction and arranged at intervals along the first direction. The first frame edges and the second frame edges intersect to form a quadrilateral. The first frame edges and the second frame edges include first surfaces and second surfaces which are oppositely disposed. The first surfaces of at least three angles of the four angles are provided with the first align mask. The first align mask is provided with a first align identifier. The first direction intersects with the second direction.

Abstract: A display substrate, a display device, a high-precision metal mask are provided. The display substrate includes first, second, third sub-pixels. In first direction, first and third sub-pixels are alternated to form first sub-pixel rows, second sub-pixels form second sub-pixel rows. In second direction, first and second sub-pixel rows are alternated, first direction is approximately perpendicular to second direction. Two first and two third sub-pixels in two adjacent rows and two adjacent columns form a 2*2 array, in the array, two first sub-pixels are in different rows and different columns, two third sub-pixels are in different rows and different columns, at least one of two first and two third sub-pixels is a pattern where corner is cut off, connection lines of centers of two first and two third sub-pixels form non-square virtual quadrilateral, and second sub-pixel is within virtual quadrilateral.

Abstract: A mask sheet for evaporation on a substrate is provided, and the mask sheet includes: a plurality of first ribs extending in a first direction; and a plurality of second ribs extending in a second direction intersecting the first direction, the mask sheet includes a first region for an evaporation of a first exposure region of the substrate and a second region for an evaporation of a second exposure region of the substrate, a third rib is provided at a boundary of the first region and the second region so as to shield an overlapping exposure region between the first exposure region and the second exposure region, the third rib includes a first sub-rib, a second sub-rib, and a spacer for separating the first sub-rib and the second sub-rib, and the third rib has a width greater than that of the first rib.

Abstract: Embodiments described herein provide a cross-lingual sentence alignment framework that is trained only on rich-resource language pairs. To obtain an accurate aligner, a pretrained multi-lingual language model is used, and a classifier is trained on parallel data from rich-resource language pairs. This trained classifier may then be used for cross-lingual transfer with low-resource languages.

Abstract: A display substrate, a display device, a high-precision metal mask are provided. The display substrate includes first, second, third sub-pixels. In first direction, first and third sub-pixels are alternated to form first sub-pixel rows, second sub-pixels form second sub-pixel rows. In second direction, first and second sub-pixel rows are alternated, first direction is approximately perpendicular to second direction. Two first and two third sub-pixels in two adjacent rows and two adjacent columns form a 2*2 array, in the array, two first sub-pixels are in different rows and different columns, two third sub-pixels are in different rows and different columns, at least one of two first and two third sub-pixels is a pattern where corner is cut off, connection lines of centers of two first and two third sub-pixels form non-square virtual quadrilateral, and second sub-pixel is within virtual quadrilateral. The display effect can be improved.

Abstract: A display substrate and a display apparatus are provided. The display substrate includes a base substrate, a crack detection line and an electrostatic discharge element electrically connected to the crack detection line. The base substrate includes a main body area, an auxiliary area, and a necked-down area connecting the main body area and the auxiliary area, and a display area of the display substrate is in the main body area. The crack detection line surrounds, at least in part, the display area and extends through the necked-down area to the auxiliary area. A length direction of the necked-down area is parallel to a first direction; the main body area, the necked-down area and the auxiliary area are connected in a second direction intersecting the first direction; and the electrostatic discharge element is located in the auxiliary area.

Abstract: A method for data processing is described. The method includes receiving an indication of a reference string from a cloud client of a content generation service. The method further includes generating multiple candidate strings associated with the reference string based on using a machine learning model to calculate similarity metrics between the reference string and the multiple candidate strings, where the machine learning model is trained using a dataset of annotated strings. The method further includes selecting a quantity of the candidate strings based on filtering the multiple candidate strings according to the similarity metrics. The method further includes causing the quantity of candidate strings to be displayed at the cloud client. The method further includes receiving feedback associated with the quantity of candidate strings and a selection of at least one candidate string displayed at the cloud client.

Abstract: The present disclosure provides a display substrate and related devices. The display substrate includes a plurality of first sub-pixels, second sub-pixels and third sub-pixels. In a first direction, the first sub-pixels and the third sub-pixels are arranged alternately to form a plurality of first sub-pixel rows, the second sub-pixels form a plurality of second sub-pixel rows, the first sub-pixel rows and the second sub-pixel rows are arranged alternately in a second direction, connection lines of center points of two first sub-pixels and two third sub-pixels form a first virtual quadrilateral, the two first sub-pixels are located at two vertex angles of the first virtual quadrilateral which are opposite to each other, one second sub-pixel is located within the first virtual quadrilateral, and the first virtual quadrilateral includes two interior angles each being equal to 90° and two interior angles each being not equal to 90°.

Abstract: A display substrate and a display device are provided. The display substrate includes a base substrate and a plurality of subpixels provided on the base substrate. Each subpixel includes a pixel circuit. The plurality of subpixels include first subpixels. The display substrate further includes power lines. Each power line is configured to provide a first supply voltage for the corresponding first subpixels, the power line is provided on a side of the pixel electrode of the first subpixel near the base substrate, and the power line includes a power line main body and a power line protrusion protruded from the power line main body. The power line protrusion is at least partially overlapped with the pixel electrode of the first subpixel in a direction perpendicular to the base substrate. The display substrate can effectively improve the display quality.

Abstract: A display substrate and a display device are provided. The display substrate includes a base substrate and a plurality of subpixels provided on the base substrate. Each subpixel includes a pixel circuit. The plurality of subpixels include first subpixels. The display substrate further includes power lines. Each power line is configured to provide a first supply voltage for the corresponding first subpixels, the power line is provided on a side of the pixel electrode of the first subpixel near the base substrate, and the power line includes a power line main body and a power line protrusion protruded from the power line main body. The power line protrusion is at least partially overlapped with the pixel electrode of the first subpixel in a direction perpendicular to the base substrate. The display substrate can effectively improve the display quality.

Abstract: A display substrate and related devices are provided. The display substrate includes a plurality of first sub-pixels, second sub-pixels and third sub-pixels. In a first direction, the first sub-pixels and the third sub-pixels are arranged alternately to form a plurality of first sub-pixel rows, the second sub-pixels form a plurality of second sub-pixel rows, the first sub-pixel rows and the second sub-pixel rows are arranged alternately in a second direction, connection lines of center points of two first sub-pixels and two third sub-pixels form a first virtual quadrilateral, the two first sub-pixels are located at two vertex angles of the first virtual quadrilateral which are opposite to each other, one second sub-pixel is located within the first virtual quadrilateral, and the first virtual quadrilateral includes two interior angles each being equal to 90° and two interior angles each being not equal to 90°.

Abstract: A gamma tuning method, apparatus, device, and storage medium and belongs to the field of display technology. The method includes: selecting a plurality of different target dimming values, and correspondingly setting, for each of the target dimming values, a same voltage gamma max power and a same voltage gamma small power; for each of the target dimming values, obtaining a gamma register value for 0 gray scale under the target dimming value according to the voltage gamma max power, the voltage gamma small power, a maximum gamma register value, and a target voltage for 0 gray scale under the target dimming value; and writing the gamma register value for 0 gray scale under the target dimming value into a gamma register of the display panel.

Abstract: A display substrate and a display device are provided. The display substrate includes a base substrate and a plurality of subpixels provided on the base substrate. Each subpixel includes a pixel circuit. The plurality of subpixels include first subpixels. The display substrate further includes power lines. Each power line is configured to provide a first supply voltage for the corresponding first subpixels, the power line is provided on a side of the pixel electrode of the first subpixel near the base substrate, and the power line includes a power line main body and a power line protrusion protruded from the power line main body. The power line protrusion is at least partially overlapped with the pixel electrode of the first subpixel in a direction perpendicular to the base substrate. The display substrate can effectively improve the display quality.