17 articles tagged with "Machine Learning"
The newly introduced VibeThinker model, with 3 billion parameters, demonstrates enhanced reasoning capabilities compared to Opus 4.5, marking a significant step in AI advancements.
Explore a detailed guide on how to effectively run Unsloth GLM-5.2 on your local machine, including installation instructions and troubleshooting tips.
A new framework called ITNet merges convolutional, recurrent, and transformer networks into a single learnable integral transform, potentially reshaping AI research.
Z.ai has launched its GLM-5.2 model, which closely competes with Claude Opus 4.8 while exclusively using Huawei technology, presenting notable cost benefits.
A cycling enthusiast has successfully indexed 2,207 GoPro videos totaling 669 GB using an M1 Max computer and open-source machine learning models to uncover memorable moments.
A recent study highlights the growing influence of AI agents in decision-making, traditionally a human domain, and explores their potential to enhance decision support systems.
A recent study published on arXiv emphasizes the importance of AI explanations in fostering user trust, particularly in large reasoning models.
A new study published on ArXiv explores innovative techniques aimed at improving the capabilities of multi-table question answering, focusing on evidence retrieval and schema linking.
The VAMPS benchmark sheds light on the performance of multimodal large language models in mathematical problem solving, revealing both capabilities and challenges.
A recent study delves into the complexities of large language model (LLM) agents, focusing on the distinction between harness updating and harness benefit in their task execution.
A recent publication on ArXiv presents a new methodology aimed at improving generalization within agentic reasoning systems, addressing key challenges faced by large language models.
A new study proposes a subgoal-based policy tree search method aimed at enhancing efficiency in single-agent deterministic scenarios, emphasizing the importance of subgoal generation.
The Rotary GPU technology is designed to enhance the execution of large Mixture of Experts models in environments with limited VRAM, potentially transforming AI applications.
A new approach to AI attribution is introduced, challenging traditional methods and focusing on specialized component hierarchies.
EVE-Agent proposes a novel framework for self-evolving agents that emphasizes justification in their learning processes, potentially transforming AI development.
A recent study presents PathCal, a calibration method designed to improve the efficiency of Large Reasoning Language Models in complex reasoning tasks.
The recent study explores methods for constructing agent memory in AI, focusing on both offline and online strategies to enhance performance in cold-start scenarios.