T2_ranking

Model	map	mrr
bge-base-zh-v1.5	0.657	0.768
+ fine-tuning	0.701	0.816
bge-reranker-base	0.666	0.761
+ fine-tuning	0.691	0.806

1. Prepare dataset

t2-reranking data

python prepare_t2ranking_data.py

2. Finetune embedding

sh embed_pairwise_train.sh

sh embed_llm_train.sg

Rerank

sh rerank_cross_encoder.sh

sh rerank_colbert.sh

sh rerank_llm.sh

Evaluate