cross-encoder-ettin-17m-Hinge

This model is a cross-encoder based on jhu-clsp/ettin-encoder-17m. 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("jhu-clsp/ettin-encoder-17m")
model = AutoModelForSequenceClassification.from_pretrained("xpmir/cross-encoder-ettin-17m-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	28.64	34.19
trec2019	85.08	57.44
trec2020	91.83	60.57
fever	65.06	66.15
arguana	12.49	18.79
climate_fever	19.74	14.58
dbpedia	55.32	29.80
fiqa	32.74	26.19
hotpotqa	67.98	50.04
nfcorpus	42.62	24.14
nq	37.44	42.07
quora	74.49	75.53
scidocs	19.25	10.30
scifact	56.89	59.72
touche	61.38	32.74
trec_covid	79.74	63.56
robust04	56.76	33.10
lotte_writing	52.67	43.97
lotte_recreation	48.91	44.02
lotte_science	36.10	30.71
lotte_technology	40.12	32.24
lotte_lifestyle	59.40	50.55
Mean In Domain	68.52	50.73
BEIR 13	48.09	39.51
LoTTE (OOD)	48.99	39.10