Update README.md

README.md

@@ -104,6 +104,67 @@ After training the model, you can evaluate the model by running the following co
 python src/convert2trec.py output/res.step-20 && python src/msmarco_eval.py data/qrels.retrieval.dev.tsv output/res.step-20.trec && path_to/trec_eval -m ndcg_cut.5 data/qrels.dev.tsv res.step-20.trec
 ```
 
+BM25 on DEV set
+```bash
+#####################
+MRR @10: 0.35894801237316354
+QueriesRanked: 24831
+recall@1: 0.05098711868967141
+recall@1000: 0.7464097131133757
+recall@50: 0.4942572226146033
+#####################
+```
+
+DPR w/o hard negatives on DEV set
+```bash
+#####################
+MRR @10: 0.4856112079562753
+QueriesRanked: 24831
+recall@1: 0.07367235058688999
+recall@1000: 0.9082753169878586
+recall@50: 0.7099350889583964
+#####################
+```
+
+DPR w/ hard negatives on DEV set
+```bash
+#####################
+MRR @10: 0.5166915171959451
+QueriesRanked: 24831
+recall@1: 0.08047455688965123
+recall@1000: 0.9135220125786163
+recall@50: 0.7327044025157232
+#####################
+```
+
+BM25 retrieved+CE reranked on DEV set
+```bash
+#####################
+MRR @10: 0.5188107959009376
+QueriesRanked: 24831
+recall@1: 0.08545219116806242
+recall@1000: 0.7464097131133757
+recall@50: 0.595298153566744
+#####################
+ndcg_cut_20             all     0.4405
+ndcg_cut_100            all     0.4705
+#####################
+```
+
+DPR retrieved+CE reranked on DEV set
+```bash
+#####################
+MRR @10: 0.5508822816845231
+QueriesRanked: 24831
+recall@1: 0.08903406988867588
+recall@1000: 0.9135220125786163
+recall@50: 0.7393720781623112
+#####################
+ndcg_cut_20             all     0.5131
+ndcg_cut_100            all     0.5564
+#####################
+```
+
 ## License
 The dataset is licensed under the [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0.html).