Memory Fidelity Evaluation on Memory-based Experiment Seen Features
-0.068P-DiffRDG
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| RDGBackbone=GPT-4o-mini-2024-07-182025.11 | -0.068 | 0 | 0.045 | 4.3 | |
| RDGBackbone=deepseek-r12025.11 | 0.006 | -0.003 | 0.04 | 2.1 | |
| RDGBackbone=GPT-4o-2024-08-062025.11 | 0.01 | 0 | 0.035 | 1.9 | |
| CDDBackbone=deepseek-r12025.11 | 0.023 | -0.018 | 0.065 | 45.2 | |
| RDGBackbone=qwen3-235b-a22b2025.11 | 0.028 | -0.001 | 0.035 | 1.1 | |
| CoTBackbone=deepseek-r12025.11 | 0.03 | -0.053 | 0.045 | 6.9 | |
| CDDBackbone=qwen3-235b-a22b2025.11 | 0.035 | 0.001 | 0.058 | 25.3 | |
| CDDBackbone=GPT-4o-2024-08-062025.11 | 0.054 | -0.002 | 0.042 | 32.6 | |
| BaselineBackbone=deepseek-r12025.11 | 0.057 | -0.041 | 0.083 | 287.3 | |
| DDDBackbone=deepseek-r12025.11 | 0.068 | -0.048 | 0.092 | 312.6 | |
| BaselineBackbone=qwen3-235b-a22b2025.11 | 0.07 | 0.002 | 0.074 | 125.6 | |
| ExplicitBackbone=qwen3-235b-a22b2025.11 | 0.07 | 0.002 | 0.05 | 7.8 | |
| ExplicitBackbone=GPT-4o-2024-08-062025.11 | 0.072 | 0 | 0.045 | 6.6 | |
| ExplicitBackbone=deepseek-r12025.11 | 0.081 | 0.003 | 0.055 | 5.4 | |
| DDDBackbone=qwen3-235b-a22b2025.11 | 0.082 | 0.008 | 0.078 | 142.8 | |
| CoTBackbone=GPT-4o-2024-08-062025.11 | 0.088 | 0 | 0.048 | 8.8 | |
| Self-debiasBackbone=deepseek-r12025.11 | 0.1 | -0.008 | 0.062 | 9.2 | |
| BaselineBackbone=GPT-4o-2024-08-062025.11 | 0.12 | 0 | 0.05 | 151 | |
| ExplicitBackbone=GPT-4o-mini-2024-07-182025.11 | 0.123 | 0.022 | 0.048 | 9.8 | |
| Self-debiasBackbone=GPT-4o-2024-08-062025.11 | 0.132 | 0.079 | 0.055 | 11 | |
| DDDBackbone=GPT-4o-2024-08-062025.11 | 0.138 | 0.012 | 0.055 | 168.5 | |
| CDDBackbone=GPT-4o-mini-2024-07-182025.11 | 0.145 | 0.042 | 0.055 | 85.7 | |
| CounterfactualBackbone=GPT-4o-2024-08-062025.11 | 0.156 | 0.08 | 0.058 | 12.5 | |
| CoTBackbone=GPT-4o-mini-2024-07-182025.11 | 0.162 | 0.045 | 0.052 | 12.5 | |
| Self-debiasBackbone=GPT-4o-mini-2024-07-182025.11 | 0.177 | 0.06 | 0.065 | 15.3 | |
| Self-debiasBackbone=qwen3-235b-a22b2025.11 | 0.177 | 0.062 | 0.049 | 18.8 | |
| CounterfactualBackbone=GPT-4o-mini-2024-07-182025.11 | 0.217 | 0.074 | 0.072 | 18.7 | |
| CoTBackbone=qwen3-235b-a22b2025.11 | 0.282 | 0.158 | 0.051 | 29.6 | |
| BaselineBackbone=GPT-4o-mini-2024-07-182025.11 | 0.321 | 0.096 | 0.065 | 506.5 | |
| DDDBackbone=GPT-4o-mini-2024-07-182025.11 | 0.358 | 0.115 | 0.068 | 558.4 | |
| CounterfactualBackbone=qwen3-235b-a22b2025.11 | 0.397 | 0.245 | 0.051 | 31.5 | |
| CounterfactualBackbone=deepseek-r12025.11 | 0.488 | 0.499 | 0.057 | 45.8 |