Skip to content
Trending storyDeveloping

Reranker fine-tuning beats decision fine-tuning for passage reranking on Qwen models up to 9B

1 article1 sourcesince Oct 27Last article

Overview

AISummary of 1 article

A controlled arXiv study compares untuned, decision-fine-tuned, and reranker-fine-tuned Qwen backbones from 0.6B to 27B parameters on ten TREC DL and BEIR collections.

Elvis Saravia shares the paper, which reports that from 0.6B to 9B, reranker fine-tuning raises nDCG@10 by 5 to 29 points over the other two versions.

Decision fine-tuning stays within 2.5 points below and 0.7 points above the untuned model at those sizes. At 27B, the two fine-tuned versions score about the same.

Written by AI from the articles below · updated Oct 27, 4:50 PM ET

Check the sources:

Latest developmentOct 27, 4:20 PM ET
Decision models trail fine-tuned rerankers on ranking across Qwen sizes

Article timeline

The articles in this story. Times are ET.

Oct 27
  1. elvisX
    Decision models trail fine-tuned rerankers on ranking across Qwen sizes

    AIA controlled arXiv study compares untuned, decision-fine-tuned, and reranker-fine-tuned Qwen backbones from 0.6B to 27B on ten TREC DL and BEIR collections. From 0.6B to 9B, reranker fine-tuning improves nDCG@10 by 5 to 29 points over the other two versions, while decision fine-tuning stays within 2.5 points below and 0.7 points above the untuned model. At 27B the two fine-tunes score about the same.

    Image from @omarsar0's post

Heat Trend

Current comparable heat
7
Peak comparable heat
10Oct 27, 5:00 PM ET
Change in past 24 hours
–

The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.