Share your thoughts, 1 month free Claude Pro on usSee more

Narrative Retrieval on CARDS

0.505MAP

SpecFi-CS (gemma-3-27b-it abliterated + Qwen3-Embedding-4b)

Updated 2mo ago

Evaluation Results

Method	Links
SpecFi-CS (gemma-3-27b-it abliterated + Qwen3-Embedding-4b) 2026.03		0.505	0.738	0.66	0.49	24.079
static* (gpt-4o + text-embedding-3-large) 2026.03		0.488	0.713	0.649	0.487	16.006
SpecFi-CS (gemma-3-27b-it + Qwen3-Embedding-4b) 2026.03		0.468	0.709	0.631	0.492	27.033
static* (gemma-3-27b-it abliterated + Qwen3-Embedding-4b) 2026.03		0.464	0.679	0.637	0.468	40.739
SpecFi-DR (gemma-3-27b-it abliterated + Qwen3-Embedding-4b) 2026.03		0.457	0.693	0.619	0.453	34.902
static* (gemma-3-27b-it + Qwen3-Embedding-4b) 2026.03		0.435	0.635	0.616	0.435	38.284
SpecFi-CS (gpt-4o + text-embedding-3-large) 2026.03		0.426	0.66	0.597	0.456	15.124
SpecFi-DR (gemma-3-27b-it + Qwen3-Embedding-4b) 2026.03		0.424	0.63	0.581	0.432	39.321
SpecFi-DR (gpt-4o + text-embedding-3-large) 2026.03		0.421	0.682	0.6	0.44	17.19
zero shot (gpt-4o + text-embedding-3-large) 2026.03		0.321	0.509	0.487	0.37	29.091
zero shot (gemma-3-27b-it + Qwen3-Embedding-4b) 2026.03		0.313	0.469	0.456	0.371	18.867
zero shot (gemma-3-27b-it abliterated + Qwen3-Embedding-4b) 2026.03		0.295	0.428	0.436	0.308	17.205