Molecule Captioning On Chebi 20

Metrics

BLEU-2

BLEU-4

METEOR

ROUGE-1

ROUGE-2

ROUGE-L

Text2Mol

Results

Performance results of various models on this benchmark

Model Name	BLEU-2	BLEU-4	METEOR	ROUGE-1	ROUGE-2	ROUGE-L	Text2Mol	Paper Title	Repository
MolT5-Base	54.0	45.7	56.9	63.4	48.5	57.8	54.7	Translation between Molecules and Natural Language
Mol-LLM (LLaMA2-Chat)	72.70	-	-	-	-	-	-	-	-
MolReGPT (GPT-4-0314)	60.7	52.5	61.0	63.4	47.6	56.2	58.5	Empowering Molecule Discovery for Molecule-Caption Translation with Large Language Models: A ChatGPT Perspective	-
LaMolT5-Large	60.2	52.1	63.4	65.5	51.2	59.8	59.7	Automatic Annotation Augmentation Boosts Translation between Molecules and Natural Language	-
Text+Chem T5-augm-Base	62.5	54.2	64.8	68.2	54.3	62.2	-	Unifying Molecular and Textual Representations via Multi-task Language Modelling
Mol-LLM (Mistral-Instruct-v0.2)	73.2	-	-	-	-	-	-	-	-
BioT5	63.5	55.6	65.6	69.2	55.9	63.3	60.3	BioT5: Enriching Cross-modal Integration in Biology with Chemical Knowledge and Natural Language Associations
MoMu+MolT5-Large	59.9	51.5	59.7	-	-	-	58.2	A Molecular Multimodal Foundation Model Associating Molecule Graphs with Natural Language
Mol-LLM	56.0	49.0	59.3	52.4	37.0	46.7	-	Mol-LLM: Generalist Molecular LLM with Improved Graph Utilization	-
MolReGPT (GPT-3.5-turbo)	56.5	48.2	62.3	45.0	54.3	58.5	56.0	Empowering Molecule Discovery for Molecule-Caption Translation with Large Language Models: A ChatGPT Perspective	-
MolT5-Large	59.4	50.8	61.4	65.4	51.0	59.4	58.2	Translation between Molecules and Natural Language
Text+Chem T5-augm-Small	56.0	47.0	58.8	63.8	48.8	58	-	Unifying Molecular and Textual Representations via Multi-task Language Modelling
LaMolT5-Base	57.4	48.5	59.6	63.4	47.8	56.4	59.9	Automatic Annotation Augmentation Boosts Translation between Molecules and Natural Language	-
LaMolT5-Small	53.9	44.6	56.6	62.0	46.9	56.3	58.8	Automatic Annotation Augmentation Boosts Translation between Molecules and Natural Language	-
PEIT-GEN	59.8	53.4	67.6	70.0	58.2	65.3	-	Property Enhanced Instruction Tuning for Multi-task Molecule Generation with Large Language Models	-
MolFM-Base	58.5	49.8	60.7	65.3	50.8	59.4	57.6	MolFM: A Multimodal Molecular Foundation Model
MoMu+MolT5-Base	54.9	46.2	57.6	-	-	-	55.8	A Molecular Multimodal Foundation Model Associating Molecule Graphs with Natural Language
Mol-LLM (SELFIES)	58.7	51.5	61.7	62.7	48.7	57.1	-	Mol-LLM: Generalist Molecular LLM with Improved Graph Utilization	-
MoMu+MolT5-Small	53.2	44.5	55.7	-	-	-	55.3	A Molecular Multimodal Foundation Model Associating Molecule Graphs with Natural Language
MolCA, Galac125M	61.6	52.9	63.9	67.4	53.3	61.5	-	MolCA: Molecular Graph-Language Modeling with Cross-Modal Projector and Uni-Modal Adapter

0 of 32 row(s) selected.