cross-encoder-ettin-32m-MarginMSE

This model is a cross-encoder based on jhu-clsp/ettin-encoder-32m. It was trained on Ms-Marco using loss marginMSE 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 marginMSE

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-32m-MarginMSE")
model = AutoModelForSequenceClassification.from_pretrained("xpmir/cross-encoder-ettin-32m-MarginMSE")

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	36.33	42.63
trec2019	92.25	66.28
trec2020	92.35	65.97
fever	78.92	79.03
arguana	20.66	30.77
climate_fever	29.51	21.86
dbpedia	72.56	42.92
fiqa	43.96	35.50
hotpotqa	86.72	70.90
nfcorpus	49.70	29.69
nq	49.32	54.54
quora	78.89	80.74
scidocs	27.14	15.26
scifact	65.93	68.86
touche	65.74	34.12
trec_covid	92.02	69.09
robust04	65.18	43.58
lotte_writing	67.49	58.88
lotte_recreation	58.89	53.47
lotte_science	47.50	39.33
lotte_technology	54.12	45.04
lotte_lifestyle	71.58	62.12
Mean In Domain	73.64	58.29
BEIR 13	58.54	48.71
LoTTE (OOD)	60.79	50.40