cross-encoder-ettin-68m-MarginMSE

This model is a cross-encoder based on jhu-clsp/ettin-encoder-68m. 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("jhu-clsp/ettin-encoder-68m")
model = AutoModelForSequenceClassification.from_pretrained("xpmir/cross-encoder-ettin-68m-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	39.84	46.58
trec2019	95.16	74.12
trec2020	94.91	73.19
fever	83.02	82.50
arguana	22.05	32.62
climate_fever	33.42	24.99
dbpedia	77.40	47.51
fiqa	47.61	39.66
hotpotqa	89.40	73.99
nfcorpus	54.93	34.23
nq	54.50	59.29
quora	81.80	83.56
scidocs	29.49	16.60
scifact	69.25	72.31
touche	61.59	35.55
trec_covid	92.45	75.26
robust04	68.93	47.00
lotte_writing	71.97	63.08
lotte_recreation	62.67	56.86
lotte_science	49.79	41.38
lotte_technology	56.84	48.10
lotte_lifestyle	73.33	63.91
Mean In Domain	76.64	64.63
BEIR 13	61.30	52.16
LoTTE (OOD)	63.92	53.39