1 points | by teugent 13 hours ago ago
1 comments
We ran a controlled 110-cycle benchmark on GPT-4o under identical context limits and temperature.
Baseline used a static system prompt. SIGMA runtime reasserts identity and consolidates context each cycle.
Results: - 60.7% fewer tokens (1322 → 520) - 20.9% lower latency (3.22s → 2.55s) - baseline drifted at cycle 23 and collapsed by cycle 73
No fine-tuning, no RAG, no larger context window. Just runtime-level control.
Happy to answer questions.
We ran a controlled 110-cycle benchmark on GPT-4o under identical context limits and temperature.
Baseline used a static system prompt. SIGMA runtime reasserts identity and consolidates context each cycle.
Results: - 60.7% fewer tokens (1322 → 520) - 20.9% lower latency (3.22s → 2.55s) - baseline drifted at cycle 23 and collapsed by cycle 73
No fine-tuning, no RAG, no larger context window. Just runtime-level control.
Happy to answer questions.