cross-encoder-ettin-150m-BCE

This model is a cross-encoder based on jhu-clsp/ettin-encoder-150m. It was trained on Ms-Marco using loss bce 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 bce

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("xpmir/cross-encoder-ettin-150m-BCE")
model = AutoModelForSequenceClassification.from_pretrained("xpmir/cross-encoder-ettin-150m-BCE")

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	36.34	42.84
trec2019	90.84	66.43
trec2020	88.01	63.79
fever	78.31	78.70
arguana	15.15	22.54
climate_fever	26.57	19.94
dbpedia	68.87	39.94
fiqa	44.46	36.50
hotpotqa	84.70	67.70
nfcorpus	48.09	29.06
nq	48.85	54.36
quora	68.81	71.56
scidocs	25.04	14.23
scifact	66.45	68.63
touche	59.39	31.52
trec_covid	86.17	66.64
robust04	54.62	34.82
lotte_writing	69.11	60.45
lotte_recreation	59.71	55.24
lotte_science	45.64	38.20
lotte_technology	51.79	44.22
lotte_lifestyle	70.66	61.44
Mean In Domain	71.73	57.69
BEIR 13	55.45	46.26
LoTTE (OOD)	58.59	49.06