Logical Reasoning on PrOntoQA (Acc, NLDD)
100AccuracyLlama-3.1-8B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Llama-3.1-8BRegime=Faithful Regime (Causal Reasoning)2026.02 | 100 | 20.6 | |
| DeepSeek-Coder-6.7BRegime=Faithful Regime (Causal Reasoning)2026.02 | 100 | 84.3 | |
| Gemma-2-9BRegime=Anti-Faithful Regime (Post-Hoc Rationalization)2026.02 | 99 | -52.5 |