How to convert real agent tracking into reinforcement learning RL to improve policy LLM without changing the existing agent stack? The Microsoft AI team released Agent Lightning to help optimize multi-agent systems. Agent Lightning...
Can a compact late interactive retriever index once and provide accurate cross-language search with fast inference? Liquid AI releases LFM2-ColBERT-350Ma compact late interactive retriever for multilingual and cross-language search. Documents can be indexed in...
In this tutorial, we explore how exploration strategies can shape intelligent decision-making through agent-based problem solving. We built and trained three agents, namely Q-Learning with epsilon-greedy exploration, Upper Confidence Bound (UCB), and Monte Carlo...
Can an open source MoE truly support agent coding workflows at a fraction of the cost of the flagship model, while maintaining long-term tool usage across MCP, shell, browser, retrieval, and code? The MiniMax...
Can we render long text as an image and use VLM to achieve 3-4x token compression, scaling 128K context to a 1M token workload while maintaining accuracy? A team of researchers from Zhipu AI...
Pyversity is a fast, lightweight Python library designed to increase the diversity of results in retrieval systems. Retrievals often return very similar items, leading to redundancy. Pyversity effectively reorders these results to display relevant...
In this tutorial, we create a fully interactive, visually compelling data visualization dashboard using Bokeh. We first transform raw data into insightful charts, then enhance them with features like linked brushes, color gradients, and...
In this tutorial, we build an advanced agent retrieval augmented generation (RAG) system that goes beyond simple question answering. We designed it to intelligently route queries to the correct knowledge source, perform self-checks to...
Serving large language models often wastes GPU memory because the engine pre-reserves large static KV cache areas for each model, even when requests are bursty or idle. Meet ‘kvcached’‘, a library that provides virtualized,...
Large language models (LLMs) provide several parameters that allow you to fine-tune their behavior and control how they generate responses. If the model does not produce the desired output, the problem usually lies in...