Poetiq’s Meta-System Automatically Builds a Model-Agnostic Harness That Improved Every LLM Tested on LiveCodeBench Pro Without Fine-Tuning
One inference harness. Eight models tested. Every single one improved — no fine-tuning required.
Why it matters
Poetiq's meta-system demonstrates a novel approach to model-agnostic optimization that bypasses traditional fine-tuning, suggesting a new category of inference-layer improvements that work across competitive model families. This challenges the assumption that model-specific tuning is necessary for performance gains.
The key facts
8 to knowMeta-system built using only Gemini 3.1 Pro
Same harness applied to 8 models without modification
100% success rate: all models showed improvement
No fine-tuning required
No access to model internals needed
Tested on LiveCodeBench Pro benchmark
Models tested include: GPT 5.5 High, Kimi K2.6, Gemini 3.0 Flash, and 4 others
Published May 15, 2026
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: Poetiq's Meta-System automatically constructed and optimized an inference harness for LiveCodeBench Pro using only Gemini 3.1 Pro — no fine-tuning, no model internals. The same harness, applied without modification to GPT 5.5 High, Kimi K2.6, Gemini 3.0 Flash, and four other models, improved every…