TeleResilienceBench - SixG Bench: leaderboard
TeleResilienceBench (2026) on its SixG Bench subset (AI-native 6G network analysis): the model sees the question, the options and the first half of a flawed reasoning trace that led a weak generator to a wrong answer, and must continue it to the correct option. Score is the percent of the 97 instances recovered (correct flip rate); higher is better.
Metric: Correct flip rate (CFR, %): share of the 97 SixG Bench instances (AI-native 6G network analysis, GSMA Open-Telco ot-lite) where the model, given the question, the options and the first half of a flawed reasoning trace from qwen3.5:2b that ended in a wrong answer, continues the reasoning in its thinking stream and returns the correct option; one continuation prompt, Ollama builds on one RTX 4090; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 8 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | gemma4:31B (Ollama) | 39.2 | |
| 2 | gemma4:e4B (Ollama) | 35.1 | |
| 3 | gemma4:e2B (Ollama) | 34 | |
| 4 | gemma4:26B (Ollama) | 32 | |
| 5 | nemotron-3-nano:4B (Ollama) | 28.9 | |
| 6 | qwen3.5:4B (Ollama) | 27.8 | |
| 7 | qwen3.5:9B (Ollama) | 24.7 | |
| 8 | qwen3.5:27B (Ollama) | 20.6 |
Interactive version: theaggregate.ai/benchmark?slug=teleresiliencebench-sixg-bench · How It Works · Data refreshed daily, snapshot 2026-11-21.