CMU Researchers Introduce PAPRIKA: A Fine-Tuning Approach that Enables Language Models to Develop General Decision-Making Capabilities Not Confined to Particular Environment

In today’s rapidly evolving AI landscape, one persistent challenge is equipping language models with robust decision-making abilities that extend beyond single-turn interactions. Traditional large language models (LLMs) excel at generating…

Salesforce AI Proposes ViUniT (Visual Unit Testing): An AI Framework to Improve the Reliability of Visual Programs by Automatically Generating Unit Tests by Leveraging LLMs and Diffusion Models

Visual programming has emerged strongly in computer vision and AI, especially regarding image reasoning. Visual programming enables computers to create executable code that interacts with visual content to offer correct…

AutoAgent: A Fully-Automated and Highly Self-Developing Framework that Enables Users to Create and Deploy LLM Agents through Natural Language Alone

From business processes to scientific studies, AI agents can process huge datasets, streamline processes, and help in decision-making. Yet, even with all these developments, building and tailoring LLM agents is…

Microsoft AI Introduces Belief State Transformer (BST): Enhancing Goal-Conditioned Sequence Modeling with Bidirectional Context

Transformer models have transformed language modeling by enabling large-scale text generation with emergent properties. However, they struggle with tasks that require extensive planning. Researchers have explored modifications in architecture, objectives,…

Researchers from AMLab and CuspAI Introduced Erwin: A Tree-based Hierarchical Transformer for Large-scale Physical Systems

Deep learning faces difficulties when applied to large physical systems on irregular grids, especially when interactions occur over long distances or at multiple scales. Handling these complexities becomes harder as…

Alibaba Researchers Propose START: A Novel Tool-Integrated Long CoT Reasoning LLM that Significantly Enhances Reasoning Capabilities by Leveraging External Tools

Large language models have made significant strides in understanding and generating human-like text. Yet, when it comes to complex reasoning tasks—especially those that require multi-step calculations or logical analysis—they often…

Q-Filters: A Training-Free AI Method for Efficient KV Cache Compression

Large Language Models (LLMs) have significantly advanced due to the Transformer architecture, with recent models like Gemini-Pro1.5, Claude-3, GPT4, and Llama3.1 demonstrating capabilities to process hundreds of thousands of tokens.…

A Coding Guide to Sentiment Analysis of Customer Reviews Using IBM’s Open Source AI Model Granite-3B and Hugging Face Transformers

In this tutorial, we will look into how to easily perform sentiment analysis on text data using IBM’s open-source Granite 3B model integrated with Hugging Face Transformers. Sentiment analysis, a…

CASS: Injecting Object-Level Context for Advanced Open-vocabulary semantic segmentation

This paper was just accepted at CVPR 2025. In short, CASS is as an elegant solution to Object-Level Context in open-world segmentation. They outperform several training-free approaches and even surpasses…

Starter Guide For Running Large Language Models LLMs

Running large language models (LLMs) presents significant challenges due to their hardware demands, but numerous options exist to make these powerful tools accessible. Today’s landscape offers several approaches – from…