Researchers present CRINN, a reinforcement learning-based approach to optimize approximate nearest-neighbor search algorithms for improved performance in AI retrieval systems.
This paper introduces a machine learning framework to optimize the geometric design of V-beam thermal sensors under specific temperature and stress constraints.
This study evaluates the performance of various pre-trained audio models for music structure analysis without the need for supervised annotations.
MoP-JEPA is a proposed world model architecture that uses a mixture of predictors to better handle stochastic transitions in latent state prediction.
This paper proposes using the Toulmin model of argumentation to structure and interpret machine learning predictions for retinal diagnostics.
A new framework called Structured Thoughts organizes large language model reasoning into distinct blocks to enhance efficiency and context management.
This research explores optimal strategies for continually fine-tuning foundation models on resource-constrained devices by balancing compute costs against model performance.
Researchers developed a deep learning-based approach to analyze online handwriting for more objective and efficient dysgraphia detection in children.
Taiji Semiconductor Suzhou has increased its registered capital by approximately 42 percent to 1.02 billion yuan to support its semiconductor packaging and testing operations.
A self-supervised foundation model has been created to improve the efficiency and scalability of cardiac CT segmentation and clinical phenotyping.
The proposed PsPLUG plugin aims to resolve conflicts between user-specific personalization and explicit style instructions in large language models.
PiCSAR is proposed as a training-free method to improve the accuracy of reasoning models by selecting the most reliable chain of thought without requiring ground-truth labels.
This research formalizes the Feedback-Coupled Memory Systems architecture in continuous time, defining agent updates through decentralized economic principles.
A new study identifies that multi-hop reasoning failures in transformers are caused by a lack of supervision on bridge entities and proposes a method to mitigate this issue.
The FAST framework improves parallel reinforcement learning for autonomous driving by mitigating the straggler effect during environment sampling.
The paper explores preprocessing techniques to optimize query efficiency for propositional formulas represented in conjunctive normal form.
The Distributed Agent System framework is introduced to enhance fault tolerance and collaboration among embodied AI agents across device, edge, and cloud environments.
The open-source Python library SupplyNetPy enables high-fidelity modeling and discrete-event simulation of complex supply chain and inventory networks.
This study evaluates various pipeline optimizations for natural language to SQL translation to improve the efficiency and performance of lightweight models.
This research introduces a method for learning minimax-regret equilibria in adversarial team games with asymmetric information to counter deceptive opponent strategies.
A new lightweight multimodal memory system called LightMem-Ego is proposed to help wearable AI assistants store and retrieve long-term user experiences.
The AI YOU framework uses Bayesian updating and prompting to continuously update a user's personality profile across 22 dimensions for digital twin applications.
AgentCheck is a new open-source workbench designed to help developers test and mitigate failures in LLM agents that utilize Model Context Protocol tools.
The proposed Replayed-Prefix On-Policy Distillation method aims to reduce the computational costs associated with training agentic LLMs by using off-environment data.