cross-encoder-ELECTRA-infoNCE

This model is a cross-encoder based on google/electra-base-discriminator. It was trained on Ms-Marco using loss infoNCE 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 infoNCE

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("google/electra-base-discriminator")
model = AutoModelForSequenceClassification.from_pretrained("xpmir/cross-encoder-ELECTRA-infoNCE")

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	41.06	47.71
trec2019	95.35	75.02
trec2020	95.52	75.06
fever	80.23	80.36
arguana	20.60	30.72
climate_fever	29.69	22.16
dbpedia	75.94	45.49
fiqa	50.16	41.51
hotpotqa	88.46	71.00
nfcorpus	58.28	35.65
nq	55.53	60.49
quora	78.10	80.39
scidocs	29.06	16.48
scifact	67.34	70.36
touche	64.08	35.61
trec_covid	93.17	70.92
robust04	72.21	49.80
lotte_writing	72.46	63.84
lotte_recreation	63.54	57.97
lotte_science	48.91	40.97
lotte_technology	57.57	47.87
lotte_lifestyle	74.62	65.47
Mean In Domain	77.31	65.93
BEIR 13	60.82	50.86
LoTTE (OOD)	64.89	54.32