| --- |
| license: apache-2.0 |
| --- |
| **NexLM HTML_Knowledge-Agent Datasets** |
| Our NexLM HTML_Knowledge-Agent datasets optimize AI agents for web-based reasoning, structured HTML understanding, and reliable task execution. |
| |
| Train agents to interpret page structure, semantic elements, forms, tables, navigation, and dynamic UI states. |
| |
| Improve agentic workflows such as information extraction, browser automation, multi-step planning, and grounded decision-making. |
| |
| Provide HTML-aware context that helps models identify actionable elements rather than treating webpages as unstructured text. |
| |
| Support robust task completion across real-world interfaces, including e-commerce, dashboards, documentation, and web applications. |
| |
| Reduce ambiguity in web interactions through structured examples that connect user intent, DOM context, actions, and expected outcomes. |
| |
| Built for developers creating capable, context-aware agents that can reason over the web and execute HTML-based tasks accurately. |
| |
| **Layouts** |
| 1. What is HTML code? (text) |
| 2. How does code work? (text) |
| 3. What do good workflows, poetic/methodical code look like? (text) |
| 4. How to code (text) |
| 5. Code roots, foundations, commands, structures (text) |
| 6. Examples and logn horizon coding tasks (Q+A) |
|
|
|
|
| **Agentic long horizon multistep coding HTML** |
| 1. Plan |
| 2. Verify plan validity |
| 3. Begin working/complete task |
| 4. Check work/present |
|
|
|
|
| Created by **NexLM** w/ assistance with NexLM's Oden-1-Preview (XHIGH) |
| Apache 2.0 |
