List your company

Question Answering on TruthfulQA (MC1, MC2, MC3, Avg. Δ)

68.25MC2DECOVEC
39.681247.098154.51561.9319Apr 13, 2026Apr 15, 2026Apr 18, 2026Apr 21, 2026Apr 24, 2026Apr 27, 2026Apr 30, 2026

Evaluation results

MethodLinks
DECOVECModel=Llama-3-8B, Protocol=Few-Shot (Rand. Ext.)2026.0468.2549.9339.861.16
DECOVECModel=Llama-3-8B, Protocol=Few-Shot (KATE)2026.0467.5348.138.891.68
DECOVECModel=Llama-3-8B, Protocol=Few-Shot (BM25)2026.0467.148.5238.610.82
DECOVECModel=Llama-3-8B, Protocol=Few-Shot (Rand. Std.)2026.0465.7644.8736.871.96
DECOVECModel=Gemma-2-9B, Protocol=Few-Shot (KATE)2026.0463.2846.5534.171.06
GRM + CIRMBackbone=Gemma-2-9B-it2026.0463.06———
DECOVECModel=Gemma-2-9B, Protocol=Few-Shot (Rand. Ext.)2026.0463.0445.4333.640.78
GRMBackbone=Gemma-2-9B-it2026.0462.78———
DECOVECModel=Gemma-2-9B, Protocol=Few-Shot (BM25)2026.0460.9343.32321.15
DECOVECModel=Qwen2-7B, Protocol=Few-Shot (KATE)2026.0460.5643.3232.691.35
FsfairX + CIRMBackbone=Gemma-2-9B-it2026.0460.39———
DECOVECModel=Qwen2-1.5B, Protocol=Few-Shot (Rand. Ext.)2026.0460.3342.7632.241.81
FsfairXBackbone=Gemma-2-9B-it2026.0459.99———
GRM + CIRMBackbone=Llama-3-8B-Instruct2026.0459.5———
DECOVECModel=Qwen2-7B, Protocol=Few-Shot (Rand. Ext.)2026.0459.4342.3331.621.43
DECOVECModel=Qwen2-7B, Protocol=Few-Shot (BM25)2026.0459.3442.4831.31.1
FsfairXBackbone=Llama-3-8B-Instruct2026.0459.21———
FsfairX + CIRMBackbone=Llama-3-8B-Instruct2026.0458.89———
GRMBackbone=Llama-3-8B-Instruct2026.0458.5———
Zero-ShotModel=Llama-3-8B, Protocol=Zero-Shot2026.0458.0637.8330.69—
DECOVECModel=Llama-2-7B, Protocol=Few-Shot (Rand. Ext.)2026.0457.3639.6629.322.19
DECOVECModel=Llama-2-7B, Protocol=Few-Shot (BM25)2026.0457.1239.3828.662.61
DECOVECModel=Llama-2-7B, Protocol=Few-Shot (KATE)2026.0457.0639.9429.162.38
DECOVECModel=Qwen2-7B, Protocol=Few-Shot (Rand. Std.)2026.0456.3938.5429.941.98
DECOVECModel=Gemma-2-9B, Protocol=Few-Shot (Rand. Std.)2026.0456.1139.2428.551.88
DECOVECModel=Yi-6B, Protocol=Few-Shot (KATE)2026.0455.6638.6828.682.79
DECOVECModel=Yi-6B, Protocol=Few-Shot (Rand. Ext.)2026.0455.1438.8228.512.64
DECOVECModel=Llama-2-7B, Protocol=Few-Shot (Rand. Std.)2026.0454.9736.0127.922.05
DECOVECModel=Yi-6B, Protocol=Few-Shot (BM25)2026.0454.4338.6827.852.63
DECOVECModel=Qwen2-1.5B, Protocol=Few-Shot (KATE)2026.0453.8937.4127.470.82
DECOVECModel=Qwen2-1.5B, Protocol=Few-Shot (BM25)2026.0453.8436.7126.721.51
DECOVECModel=Yi-6B, Protocol=Few-Shot (Rand. Std.)2026.0452.0134.1826.671.59
DECOVECModel=Qwen2-0.5B, Protocol=Few-Shot (KATE)2026.0450.4334.624.981.37
Zero-ShotModel=Qwen2-7B, Protocol=Zero-Shot2026.0450.1332.4926.25—
DECOVECModel=Qwen2-1.5B, Protocol=Few-Shot (Rand. Std.)2026.0449.9133.4725.020.64
DECOVECModel=Qwen2-0.5B, Protocol=Few-Shot (Rand. Ext.)2026.0448.0132.2123.731.56
DECOVECModel=Qwen2-0.5B, Protocol=Few-Shot (BM25)2026.0447.8332.3523.091.43
DECOVECModel=Qwen2-0.5B, Protocol=Few-Shot (Rand. Std.)2026.0447.5432.3523.461.95
Zero-ShotModel=Gemma-2-9B, Protocol=Zero-Shot2026.0445.062722.08—
Zero-ShotModel=Llama-2-7B, Protocol=Zero-Shot2026.0444.9426.5821.76—
Zero-ShotModel=Qwen2-1.5B, Protocol=Zero-Shot2026.0444.726.320.78—
Zero-ShotModel=Yi-6B, Protocol=Zero-Shot2026.0444.6425.4621.5—
Zero-ShotModel=Qwen2-0.5B, Protocol=Zero-Shot2026.0440.7825.1819.55—
KICK-SETTLEDefense=SDD, Model=Llama-2-7B, Attack Type=Kick-settle2026.05———0.042
KICK-SETTLEDefense=SDD, Model=Qwen3-8B, Attack Type=Kick-settle2026.05———0.015
KICK-SETTLEDefense=SDD, Model=Llama-3.1-8B, Attack Type=Kick-settle2026.05———0.063