cross-encoder-ettin-68m-Hinge

This model is a cross-encoder based on jhu-clsp/ettin-encoder-68m. 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("xpmir/cross-encoder-ettin-68m-Hinge")
model = AutoModelForSequenceClassification.from_pretrained("xpmir/cross-encoder-ettin-68m-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	36.88	43.27
trec2019	95.64	70.28
trec2020	93.21	70.02
fever	74.34	74.86
arguana	14.82	21.93
climate_fever	18.28	13.60
dbpedia	69.03	40.06
fiqa	44.21	36.51
hotpotqa	76.58	59.91
nfcorpus	54.14	32.81
nq	48.47	53.56
quora	69.78	72.38
scidocs	26.53	14.75
scifact	66.99	69.92
touche	60.23	33.28
trec_covid	88.60	73.70
robust04	65.30	41.88
lotte_writing	70.41	60.85
lotte_recreation	59.08	53.95
lotte_science	49.30	40.56
lotte_technology	53.55	45.06
lotte_lifestyle	69.13	60.08
Mean In Domain	75.24	61.19
BEIR 13	54.77	45.94
LoTTE (OOD)	61.13	50.40