Quality Validation on Constructed high-redundancy corpus 71.98% byte reduction
4MAT Pre-Audit Issues CountLlama 3.3 70B
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Llama 3.3 70BInference Platform=Groq2026.05 | 4 | 5.03 | 0 | 198 | 1.9 | — | |
| Gemini 2.5 Flash2026.05 | 6 | 6.39 | 3 | 199 | 4.34 | — | |
| Claude Sonnet 4.62026.05 | 6 | 6.39 | 1 | 200 | 2.78 | — | |
| GPT-5.12026.05 | 7 | 7.05 | 2 | 198 | 3.61 | — |