cross-encoder-DeBERTav3-MarginMSE

This model is a cross-encoder based on microsoft/deberta-v3-base. It was trained on Ms-Marco using loss marginMSE as part of a reproducibility paper for training cross encoders: "Reproducing and Comparing Distillation Techniques for Cross-Encoders", see the paper for more details.

Contents

• Model Description
• Usage
• Evals

Model Description

This model is intended for re-ranking the top results returned by a retrieval system (like BM25, Bi-Encoders or SPLADE).

• Training Data: MS MARCO Passage
• Language: English
• Loss marginMSE

Training can be easily reproduced using the assiciated repository. The exact training configuration used for this model is also detailed in config.yaml.

Usage

Quick Start:

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

tokenizer = AutoTokenizer.from_pretrained("microsoft/deberta-v3-base")
model = AutoModelForSequenceClassification.from_pretrained("xpmir/cross-encoder-DeBERTav3-MarginMSE")

features = tokenizer("What is experimaestro ?", "Experimaestro is a powerful framework for ML experiments management...", padding=True, truncation=True, return_tensors="pt")

model.eval()
with torch.no_grad():
    scores = model(**features).logits
    print(scores)

Evaluations

We provide evaluations of this cross-encoder re-ranking the top 1000 documents retrieved by naver/splade-v3-distilbert.

dataset	RR@10	nDCG@10
msmarco_dev	38.39	45.04
trec2019	92.44	70.77
trec2020	93.24	69.48
fever	81.01	80.55
arguana	15.49	22.96
climate_fever	25.56	19.95
dbpedia	73.59	44.10
fiqa	48.06	39.70
hotpotqa	85.99	69.71
nfcorpus	49.27	29.33
nq	54.72	59.88
quora	73.21	75.37
scidocs	26.94	15.35
scifact	62.87	65.14
touche	62.05	35.65
trec_covid	95.22	76.93
robust04	66.26	45.22
lotte_writing	71.44	62.85
lotte_recreation	63.05	58.11
lotte_science	49.83	41.58
lotte_technology	58.85	48.99
lotte_lifestyle	77.11	66.26
Mean In Domain	74.69	61.76
BEIR 13	58.00	48.82
LoTTE (OOD)	64.42	53.84