cross-encoder-DeBERTav3-ADR-MSE

This model is a cross-encoder based on microsoft/deberta-v3-base. It was trained on Ms-Marco using loss ADR 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 ADR

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-ADR-MSE")

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	35.11	41.61
trec2019	93.60	74.47
trec2020	95.06	71.98
fever	82.73	82.03
arguana	14.10	21.77
climate_fever	27.96	20.23
dbpedia	73.43	42.60
fiqa	44.04	36.59
hotpotqa	75.64	58.04
nfcorpus	51.21	30.66
nq	53.19	58.19
quora	71.59	74.08
scidocs	26.33	14.86
scifact	67.32	70.28
touche	64.97	35.00
trec_covid	94.33	78.61
robust04	70.72	46.79
lotte_writing	67.14	58.43
lotte_recreation	61.62	56.57
lotte_science	48.52	40.41
lotte_technology	55.36	46.19
lotte_lifestyle	74.52	65.28
Mean In Domain	74.59	62.69
BEIR 13	57.45	47.92
LoTTE (OOD)	62.98	52.28