pearl-benchmark-models

raghvendramall/pearl-benchmark-models

pearl-benchmark-models is a machine learning model from raghvendramall. Use it for the machine learning task on the model card, and read the license before you ship it in a product.

Parameter-Efficient Adaptation with Retrieval-Augmented Learning for Molecular Property Prediction. Each dataset folder below holds its top-5 best-performing methods, ranked by the dataset's primary metric.

Downloads · 30 days

Access

Public

Updated Aug 28, 2026

Repo size

9.6 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.index8.2 GB · 85%

At a glance

Access: Public
Created: Aug 11, 2026
Updated: Aug 28, 2026
SHA: 4d27aef7

Model card

From the Hugging Face model README

PEARL Benchmark -- Top-5 Models per Dataset

Code: https://github.com/raghvendra5688/PEARL

BACE (classification, ranked by MCC)

Rank	Method	MCC	Configuration	Path
1	E2E LoRA -- Uni-Mol	0.6234	Weighted Loss	BACE/rank1_e2e-lora-uni-mol
2	E2E LoRA -- HF	0.5660	ibm__MoLFormer__XL__both__10pct, fl	BACE/rank2_e2e-lora-hf
3	FT Embed	0.5570	MolFormer_Finetuned_FL	BACE/rank3_ft-embed
4	RAFE	0.5570	MolFormer_Finetuned_FL	BACE/rank4_rafe
5	FT Embed+PC	0.5500	Molformer_Finetuned_WL_embeddings	BACE/rank5_ft-embed-pc

BBBP (classification, ranked by MCC)

Rank	Method	MCC	Configuration	Path
1	RAFE	0.7980	MolFormer_Finetuned_FL	BBBP/rank1_rafe
2	FT Embed+PC	0.7890	MolFormer_Finetuned_FL_embeddings	BBBP/rank2_ft-embed-pc
3	FT Embed	0.7870	MolFormer_Finetuned_FL	BBBP/rank3_ft-embed
4	E2E LoRA -- Uni-Mol	0.7740	Weighted Loss	BBBP/rank4_e2e-lora-uni-mol
5	E2E LoRA -- HF	0.7670	ibm__MoLFormer__XL__both__10pct, fl	BBBP/rank5_e2e-lora-hf

FLAVOR (classification, ranked by MCC)

Rank	Method	MCC	Configuration	Path
1	Chemprop	0.8090	D-MPNN	FLAVOR/rank1_chemprop
2	PC-only	0.8080	XGBoost	FLAVOR/rank2_pc-only
3	E2E LoRA -- HF	0.8020	DeepChem__ChemBERTa__77M__MTR, wl	FLAVOR/rank3_e2e-lora-hf
4	FT Embed+PC	0.8010	Molformer_Finetuned_WL_embeddings	FLAVOR/rank4_ft-embed-pc
5	RAFE	0.8000	ChemBERTa_77M_MTR_FL	FLAVOR/rank5_rafe

HERG (classification, ranked by MCC)

Rank	Method	MCC	Configuration	Path
1	E2E LoRA -- Uni-Mol	0.6093	Weighted Loss	HERG/rank1_e2e-lora-uni-mol
2	PC-only	0.6040	XGBoost	HERG/rank2_pc-only
3	FT Embed+PC	0.6015	UniMol_FL	HERG/rank3_ft-embed-pc
4	RAFE	0.5889	UniMol_FL	HERG/rank4_rafe
5	FT Embed	0.5798	UniMol_FL	HERG/rank5_ft-embed

DILI (classification, ranked by MCC)

Rank	Method	MCC	Configuration	Path
1	RAFE	0.7897	UniMol_FL	DILI/rank1_rafe
2	E2E LoRA -- HF	0.7704	DeepChem__ChemBERTa__77M__MTR, wl	DILI/rank2_e2e-lora-hf
3	PC-only	0.7650	CatBoost	DILI/rank3_pc-only
4	GCN	0.7280	GINConv	DILI/rank4_gcn
5	E2E LoRA -- Uni-Mol	0.7079	Focal Loss	DILI/rank5_e2e-lora-uni-mol

CACO2 (regression, ranked by Spearman)

Rank	Method	Spearman	Configuration	Path
1	PC-only	0.8534	CatBoost	CACO2/rank1_pc-only
2	RAFE	0.8347	ChemBERTa_77M_MTR_Huber	CACO2/rank2_rafe
3	FT Embed+PC	0.8271	ChemBERTa_77M_MTR_Huber	CACO2/rank3_ft-embed-pc
4	FT Embed	0.8211	ChemBERTa_77M_MTR_Huber	CACO2/rank4_ft-embed
5	Chemprop	0.8022	D-MPNN	CACO2/rank5_chemprop

HALF_LIFE (regression, ranked by Spearman)

Rank	Method	Spearman	Configuration	Path
1	FT Embed	0.6226	MolFormer_Finetuned_Huber	HALF_LIFE/rank1_ft-embed
2	FT Embed+PC	0.5927	MolFormer_Finetuned_Huber	HALF_LIFE/rank2_ft-embed-pc
3	RAFE	0.5684	MolFormer_Finetuned_Huber	HALF_LIFE/rank3_rafe
4	E2E LoRA -- HF	0.5406	DeepChem__ChemBERTa__77M__MTR, huber	HALF_LIFE/rank4_e2e-lora-hf
5	PC-only	0.5211	CatBoost	HALF_LIFE/rank5_pc-only

Created: Aug 11, 2026
Updated: Aug 28, 2026