Factual Knowledge Evaluation on PopQA

18Accuracy

Llama 2

Updated 4d ago

Evaluation Results

Method
Llama 2 2026.01	18	-	-	-
OLMo 2 2026.01	16	-	-	-
Unfiltered 2026.01	11	-	-	-
Alignment Upsampled 2026.01	11	-	-	-
Filtered + Alignment Upsampled 2026.01	11	-	-	-
OLMo 3 2026.01	11	-	-	-
Filtered 2026.01	10	-	-	-
Alignment Upsampled 2026.01	10	-	-	-
Alignment Upsampled 2026.01	10	-	-	-
Misalignment Upsampled 2026.01	10	-	-	-
Misalignment Upsampled 2026.01	10	-	-	-
Filtered + Alignment Upsampled 2026.01	9	-	-	-
Filtered + Alignment Upsampled 2026.01	9	-	-	-
Misalignment Upsampled 2026.01	9	-	-	-
PretrainRL 2026.02	0.5016	0.5138	0.3863	0.3067
PretrainRL 2026.02	0.4861	0.7492	0.5151	0.4048
Iterative RPO 2026.02	0.4123	0.4813	0.3228	0.4083
Continual Training (CT) 2026.02	0.3853	0.6917	0.4193	0.275
PretrainRL 2026.02	0.3769	0.6639	0.4051	0.3142
Iterative RPO 2026.02	0.3383	0.3263	0.2264	0.2817
Iterative RPO 2026.02	0.325	0.493	0.2512	0.2077
Continual Training (CT) 2026.02	0.2997	0.6036	0.3294	0.2092
Prompt (CoT) 2026.02	0.2428	0.4968	0.2706	0.2132
Llama3-8B 2026.02	0.2366	0.4869	0.2635	0.231
Qwen3-8B 2026.02	0.1632	0.3706	0.186	0.186
DPO 2026.02	0.1568	0.2602	0.1693	0.4228
Prompt (CoT) 2026.02	0.1554	0.3757	0.1797	0.1214
Continual Training (CT) 2026.02	0.134	0.3318	0.1653	0.201
DPO 2026.02	0.1308	0.2217	0.1392	0.4211
Qwen3-4B 2026.02	0.1305	0.316	0.1517	0.1238
Prompt (CoT) 2026.02	0.1092	0.3247	0.1478	0.1092
DPO 2026.02	0.0967	0.1806	0.1102	0.3388