cross-encoder-ettin-32m-DistillRankNET

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

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

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	29.69	35.29
trec2019	91.86	62.04
trec2020	85.57	63.47
fever	70.41	71.33
arguana	8.61	13.20
climate_fever	16.04	11.98
dbpedia	61.21	34.43
fiqa	32.94	25.37
hotpotqa	74.34	57.33
nfcorpus	40.43	23.10
nq	38.18	42.81
quora	72.61	73.97
scidocs	21.50	11.66
scifact	51.45	54.28
touche	64.88	31.23
trec_covid	88.83	64.72
robust04	52.38	31.19
lotte_writing	59.75	50.70
lotte_recreation	48.66	43.92
lotte_science	38.10	32.33
lotte_technology	42.30	34.81
lotte_lifestyle	59.83	50.72
Mean In Domain	69.04	53.60
BEIR 13	49.34	39.65
LoTTE (OOD)	50.17	40.61