In the news
An RLM agent for searching Wikidata, and the traces to train on it
Pleias · Published · 3 min read
In 30 seconds
- What happened
- Researchers released an open recursive language model harness enabling models to search Wikidata through code rather than pasting results into context.
- Why it matters
- Matters for engineers building knowledge assistants, semantic search systems, or tools that need to query structured databases with complex multi-step reasoning.
- Watch out
- The evaluation uses only 100 test questions; multi-hop performance remains challenging with best scores of 28 out of 50 correct answers on harder questions.
- agent
- language model
- eval
- open-weight
- gpt
The patterns behind this
- Context Editing & Tool-Result Clearing
- Agentic Context Engineering (Evolving Playbook)
- Structure-Aware Codebase Retrieval (Repo Map)
Each one covers how the technique works, when it earns its cost, and where it breaks.
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.