cross-encoder-DeBERTav3-Hinge

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

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-Hinge")

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	36.22	42.66
trec2019	96.12	70.14
trec2020	91.21	67.87
fever	75.48	75.39
arguana	14.36	21.38
climate_fever	24.22	18.07
dbpedia	67.99	38.75
fiqa	46.11	37.81
hotpotqa	75.45	57.94
nfcorpus	49.05	28.70
nq	50.32	55.25
quora	61.89	64.80
scidocs	26.07	14.97
scifact	66.01	68.71
touche	56.51	33.08
trec_covid	91.57	73.45
robust04	64.27	42.59
lotte_writing	66.27	57.80
lotte_recreation	61.18	55.67
lotte_science	46.46	38.88
lotte_technology	54.78	45.97
lotte_lifestyle	73.88	64.68
Mean In Domain	74.52	60.22
BEIR 13	54.23	45.25
LoTTE (OOD)	61.14	50.93