How do we design AI systems that can plan, reason, and act over long sequences of decisions without constant human guidance? Moonshot AI has released Kimi K2 Thinking, an open source thinking agent model that exposes the full reasoning stream of the Kimi K2 Mixture of Experts architecture. It targets workloads that need deep reasoning,…
In this tutorial, we build a Wet-Lab Protocol Planner & Validator that acts as an intelligent agent for experimental design and execution. We design the system using Python and integrate Salesforce’s CodeGen-350M-mono model for natural language reasoning. We structure the pipeline into modular components: ProtocolParser for extracting structured data, such as steps, durations, and temperatures,…
How do you turn a vague business style question over messy folders of CSV, JSON and text into reliable Python code without a human analyst in the loop? Google researchers introduce DS STAR (Data Science Agent via Iterative Planning and Verification), a multi agent framework that turns open ended data science questions into executable Python…
Most LLM agents are tuned to maximize task success. They resolve GitHub issues or answer deep research queries, but they do not reason carefully about when to ask the user questions or how to respect different interaction preferences. How can we design LLM agents that know when to ask better questions and adapt their behavior…
How do you build a single model that can learn physical skills from chaotic real world robot data without relying on simulation? Generalist AI has unveiled GEN-θ, a family of embodied foundation models trained directly on high fidelity raw physical interaction data instead of internet video or simulation. The system is built to establish scaling…
How can we reliably test whether large language models actually understand Indian languages and culture in real world contexts? OpenAI has released IndQA, a benchmark that evaluates how well AI models understand and reason about questions that matter in Indian languages across cultural domains. Why IndQA? OpenAI states that about 80 percent of people worldwide…
In this tutorial, we explore how an agent can internalize planning, memory, and tool use within a single neural model rather than relying on external orchestration. We design a compact, model-native agent that learns to perform arithmetic reasoning tasks through reinforcement learning. By combining a stage-aware actor-critic network with a curriculum of increasingly complex environments,…
How can consistency training help language models resist sycophantic prompts and jailbreak style attacks while keeping their capabilities intact? Large language models often answer safely on a plain prompt, then change behavior when the same task is wrapped with flattery or role play. DeepMind researchers propose consistent training in a simple training lens for this…
In this tutorial, we explore Hydra, an advanced configuration management framework originally developed and open-sourced by Meta Research. We begin by defining structured configurations using Python dataclasses, which allows us to manage experiment parameters in a clean, modular, and reproducible manner. As we move through the tutorial, we compose configurations, apply runtime overrides, and simulate…
Code-oriented large language models moved from autocomplete to software engineering systems. In 2025, leading models must fix real GitHub issues, refactor multi-repo backends, write tests, and run as agents over long context windows. The main question for teams is not “can it code” but which model fits which constraints. Here are seven models (and systems…
