In the news
Same Query, Different Answer: Run-to-Run Variance in Semantic Query Processing
arXiv cs.AI · Published · 3 min read
In 30 seconds
- What happened
- Researchers found that LLM-based semantic query engines produce inconsistent results across repeated runs, with 0.07% to 10.2% of individual rows changing outputs even at temperature zero.
- Why it matters
- Database engineers and teams building or evaluating semantic query systems need to know this affects benchmark reliability and real-world query consistency.
- Watch out
- The study tested existing benchmarks but did not propose a complete solution; voting and cascades reduce but do not eliminate variance entirely.
- llm
- language model
- benchmark
The patterns behind this
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.