Abstract: Methods and systems include annotating a set of training data to indicate tokens that are sensitive. Instructions are generated based on the training data, including original token sequences and respective substituted token sequences. A language model is fine-tuned using the instructions with a penalty-based loss function to generate a privacy-protected language model.

Abstract: Systems and methods for predicting an occurrence of a medical event for a patient using a trained neural network. Historical patient data is preprocessed to generate normalized training samples, and the normalized training samples are sent to a personalized deep convolutional neural network for model pretraining and updating of model parameters. The pretrained model is stored in a remote server for utilization by a local machine for personalization during a preparation time period for a medical treatment. A normalized finetuning set is generated as output, and the model parameters are iteratively finetuned. A personal prediction score for future medical events is generated, and an operation of a medical treatment device is controlled responsive to the prediction score.

Abstract: In a shift register of a driver circuit of a display panel, a drive control circuit is electrically connected to a signal input terminal, a first node, and a second node and controls a signal of the first node and a signal of the second node in response to an input signal of the signal input terminal; an output circuit is electrically connected to the first node, the second node, a first level terminal, a first clock terminal, and a signal output terminal and controls an output signal of the signal output terminal according to the signal of the first node, the signal of the second node, a first level signal of the first level terminal, and a first clock signal of the first clock terminal; the signal output terminal of an x-th stage shift register is electrically connected to a signal input terminal of a y-th stage shift register.

Abstract: Provided are a display panel and a preparation method therefor and a display device. The display panel includes a base substrate, a first planarization layer, a first metal layer, a photosensitive element, an etching stop layer, and a light-emitting element. The photosensitive element includes a first electrode and a semiconductor layer that are electrically connected to each other, the first electrode is located in the first metal layer, and the semiconductor layer is located on one side of the etching stop layer facing away from the first metal layer. In the display region, the etching stop layer includes at least one first opening. Along a direction perpendicular to the plane where the base substrate is located, a first opening does not overlap with the first metal layer.

Abstract: An RRAM structure includes a dielectric layer. A bottom electrode, a resistive switching layer and a top electrode are disposed from bottom to top on the dielectric layer. A spacer is disposed at sidewalls of the bottom electrode, the resistive switching layer and the top electrode. The spacer includes an L-shaped spacer and a sail-shaped spacer. The L-shaped spacer contacts the sidewall of the bottom electrode, the sidewall of the resistive switching layer and the sidewall of the top electrode. The sail-shaped spacer is disposed on the L-shaped spacer. A metal line is disposed on the top electrode and contacts the top electrode and the spacer.

Abstract: Systems and methods for detecting machine-generated texts by detecting motifs. The system and method including tokenizing documents into token sequences, generating at least one graph where tokens and documents are nodes, wherein the tokens are one-hot features and the documents are zero features, and wherein the graph has at least two types of connections, at least one type connects documents to tokens that the documents contain, and at least one other type links tokens that frequently appear together across all documents, classifying each node in the at least one graph using a graph neural network (GNN), and generating an optimal explanation subgraph from the classified graph, wherein the optimal explanation subgraph represents a motif. The optimal explanation subgraph can later be used to support artificial intelligence (AI) agent decision-making.

Abstract: Methods and systems include inferring a first prediction and an explanation based on a time series input and a text input using a multi-modal prototype-based encoder. A second prediction is inferred based on the text input and the explanation using a large language model. The first prediction and the second prediction are fused to generate a fused prediction. A reflection is generated based on the fused prediction, the second prediction, and the text input. The text input is refined based on the reflection.

Abstract: According to an aspect of the present invention, a system method is provided for solution decomposition, the system and method comprising in response to a query, decomposing a solution using a recursive subprocess including generating complete solutions using a policy until there are enough complete solutions to determine a reward model, selecting at least one target step of the complete solutions based on the reward model, partitioning the at least one target step into at least two segments based on a predefined partition fraction, wherein for each segment, a priority metric is estimated, wherein the at least one target step is a first step or a segment of a previously partitioned step, appending the partitioned segment with a lower priority metric, determining whether an end condition has been met, and ending the recursive subprocess if the end condition is met; and outputting a decomposition.

Abstract: Systems and methods for Evidence-based Sound Event Early Detection is provided. The system/method includes parsing collected labeled audio corpus data and real time audio streaming data utilizing mel-spectrogram, encoding features of the parsed mel-spectrograms using a trained neural network, and generating a final predicted result for a sound event based on the belief, disbelief and uncertainty outputs from the encoded mel-spectrograms.

Abstract: Disclosed are an anti-peeping film and a display device. In the present disclosure, an angle between long axes of liquid crystal molecules in a liquid crystal light-adjusting layer and planes in which polarizers are located is set to be an acute angle in a first mode, so that there is an angle between effective long axes of the liquid crystal molecules in the liquid crystal light-adjusting layer and a direction of polarized light in a side-view direction. Therefore, light in the side-view direction of the anti-peeping film is darker. Thus, an anti-peeping effect can be realized, while sharing can be realized in a non-first mode.

Abstract: A computer-implemented method for meta-learning is provided. The method includes receiving a training time series and labels corresponding to some of the training time series. The method further includes optimizing time series augmentations of the training time series using a time series augmentation selection process performed by a meta learner to obtain a selected augmentation from a plurality of candidate augmentations. The method also includes training a time series encoder with contrastive loss using the selected augmentation to obtain a learned time series encoder. The method additionally includes learning, by the learned time series encoder, a vector representation of another time series. The method further includes performing, by the learned time series encoder, a downstream task of label classification for at least a portion of the other time series.

Abstract: Methods and systems for training and deploying a neural network mode include training a modular encoder model using training data collected from heterogeneous system types. The modular encoder model includes layers of neural network blocks and a selectively enabled connections between neural network blocks of adjacent layers. Each neural network block includes neural network layers. The modular encoder model is deployed to a system corresponding to one of the heterogeneous system types.

Abstract: Provided is a bispecific chimeric antigen receptor targeting CD19 and CD22, which comprises extracellular antigen binding domains of heavy-chain variable regions and light-chain variable regions of anti-CD19 and anti-CD22 antibodies. Further provided is a bispecific CAR-T cell targeting CD19 and CD22.

Abstract: A polymer liquid crystal dimming box, a manufacturing method thereof, and a display module are provided. The polymer liquid crystal dimming box includes a first substrate, a second substrate, a polymer liquid crystal layer, a frame adhesive layer and a light-shielding structure. The light-shielding structure is disposed between the frame adhesive layer and the polymer liquid crystal layer and surrounds the polymer liquid crystal layer. The light-shielding structure is used to absorb ultraviolet light obliquely incident during the curing process of the frame adhesive layer. In this way, the problem that the polymer liquid crystal layer is cured in advance due to ultraviolet light irradiation can be avoided, the edge failure of the polymer liquid crystal dimming box can be improved, which is conducive to mass production.

Abstract: Systems and methods for detecting artificial intelligence (AI) generated computer code. Lines of code can be masked from a candidate code to obtain perturbed codes. Missing code can be generated from the perturbed codes by employing an AI code generator model to obtain machine-filled codes. Probabilities of the candidate code probability and the machine-filled codes as AI-generated can be predicted by employing a surrogate model. The candidate code can be distinguished as AI-generated by comparing the probabilities against a detection threshold to obtain detection results.

Abstract: The present disclosure provides a semiconductor epitaxial structure, including a substrate, a first-type semiconductor layer, an active region comprising multiple quantum layers stacked in sequence along a first direction, and a stress release layer between two adjacent quantum layers, wherein the first direction is perpendicular to the substrate and is directed from the substrate to the first-type semiconductor layer, and a second-type semiconductor layer sequentially stacked on a surface of the substrate; wherein the quantum layers each comprises barrier layers and potential well layers, and the barrier layers are alternately stacked with the potential well layers, and wherein at least one of the quantum layers further comprises a growth temperature transition layer between a barrier layer and a potential well layer, or an electron confinement layer between a barrier layer and a potential well layer.

Abstract: Methods and systems include embedding an input query, including contextual information. Performance and cost of executing the input query are predicted on each of a set of language models. The prediction is performed using a multi-armed bandit approach with each of the language models being represented by a respective arm. The input query is executed on a selected model that has a best balance of performance and cost.

Abstract: The present application discloses a method for analyzing a layout pattern density, comprising: step 1, providing layouts of a chip, and merging the layouts to form a wafer level layout, wherein the wafer level layout presents a first circle in a top view, and the layout comprises a plurality of mask layers; step 2, segmenting the first circle to form a plurality of check windows; step 3, searching for the mask layer containing the patterns having a height morphology, and combining the found mask layers into a pattern layer combination; step 4, sequentially calculating a pattern density of the pattern layer combination in each check window; and step 5, recording the pattern density in each check window on a third circle to form a wafer level pattern density distribution diagram. The present application can predict a height morphology of a top surface of a wafer related to a layout.

Abstract: Systems and methods for detecting artificial intelligence (AI) generated text. A candidate text can be truncated to obtain a prefix text and a remainder text by employing a truncation module. Regenerated model texts can be regenerated by utilizing the prefix text by employing an AI text generation model. Detection results can be predicted by comparing n-gram similarities of the regenerated model texts and the remainder text. The candidate text can be distinguished as AI generated text by providing explanation texts based on the detection results.