cross-encoder-DeBERTav3-DistillRankNET

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

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-DistillRankNET")

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	35.30	41.91
trec2019	94.65	74.18
trec2020	93.58	70.05
fever	82.83	81.97
arguana	13.59	21.15
climate_fever	26.82	19.27
dbpedia	72.24	42.56
fiqa	42.94	35.84
hotpotqa	78.51	60.35
nfcorpus	47.19	28.16
nq	52.10	57.12
quora	71.72	74.00
scidocs	25.04	14.36
scifact	63.12	65.74
touche	68.90	34.59
trec_covid	89.07	76.15
robust04	70.29	46.92
lotte_writing	67.04	57.94
lotte_recreation	61.21	55.99
lotte_science	48.10	40.20
lotte_technology	55.99	46.36
lotte_lifestyle	74.51	64.85
Mean In Domain	74.51	62.05
BEIR 13	56.47	47.02
LoTTE (OOD)	62.86	52.04