Model	AVE	PRP	PM	SA	SR	MPC	PSI	QPR	AP	AG

	F1*	Macro F1	F1	Macro F1	HR@1	Accuracy	F1	NDCG	F1	F_BERT
GPT-4 Turbo	0.495	0.326	0.753	0.516	0.387	0.611	0.195	0.875	0.649	0.858
Gemini Pro	0.396	0.136	0.867	0.470	0.269	0.584	0.248	0.821	0.506	0.855
Claude 2.1	0.381	0.275	0.523	0.415	0.066	0.655	0.273	0.821	0.280	0.841
Llama-2 13B-chat	0.002	0.333	0.434	0.188	0.056	0.504	0.252	0.815	0.623	0.811
Mistral-7B Instruct-v0.2	0.369	0.324	0.613	0.470	0.164	0.529	0.305	0.842	0.588	0.853

EcomGPT	0.000	0.091	0.648	0.188	0.042	0.540	0.170	0.000	0.086	0.669

SoTA task-specific model	0.546	0.588	0.995	0.573	0.265	0.703	0.389	0.859	0.830	0.858
eCeLLM-L	0.582	0.611	0.995	0.648	0.526	0.684	0.501	0.870	0.851	0.841
eCeLLM-M	0.662	0.558	0.995	0.639	0.542	0.696	0.305	0.876	0.846	0.842
eCeLLM-S	0.509	0.518	0.991	0.596	0.479	0.650	0.392	0.870	0.846	0.842
improvement (%, avg: 10.7)	21.2	3.9	0.0	13.1	40.1	-1.0	28.8	0.1	2.5	-1.9