Search results

Major upgrade for torch - these-alexia

critical

Status: Open.

#39 In artefactory/these-alexia;

· Cedric-Magnan opened on Jun 14, 2026

Add queries, paraphrased docs and pseudo-queries in the training of the codebooks to generate docids

Status: Open.

#38 In artefactory/these-alexia;

· AlexiaALLAL opened on Jun 11, 2026

DocID Analysis: semantic correlation with Kendall instead of Pearson/spearman

Status: Open.

#34 In artefactory/these-alexia;

· AlexiaALLAL opened on May 29, 2026

distance hamming instead of cosinenfor docid analysis with faiss

Status: Open.

#33 In artefactory/these-alexia;

· AlexiaALLAL opened on Apr 21, 2026

evaluate clusters with scikit-learn metrics like silhouette score, calinski-harabasz index, davies-bouldin index, etc.

Status: Open.

#27 In artefactory/these-alexia;

· AlexiaALLAL opened on Apr 14, 2026

Deffered quantization : https://arxiv.org/pdf/2603.17052

Status: Open.

#19 In artefactory/these-alexia;

· AlexiaALLAL opened on Mar 25, 2026

Try to add weights in the argmin of the kmeans

Status: Open.

#16 In artefactory/these-alexia;

· AlexiaALLAL opened on Mar 17, 2026

Régler problème de mauvais entrainement quand j'utilise un encodeur

Status: Open.

#12 In artefactory/these-alexia;

· AlexiaALLAL opened on Mar 16, 2026

Tester encodeur sylvain : générer k embeddings, et appliquer VQ sur chaque vecteur

Status: Open.

#11 In artefactory/these-alexia;

· AlexiaALLAL opened on Mar 16, 2026

In IR loss, select hard negatives instead of random negatives

Status: Open.

#10 In artefactory/these-alexia;

· AlexiaALLAL opened on Mar 16, 2026

See if step 3 works better with RQ-IR or RQ-kmeans, PQ

Status: Open.

#6 In artefactory/these-alexia;

· AlexiaALLAL opened on Feb 23, 2026

Test IR Loss variations for docids generation

Status: Open.

#4 In artefactory/these-alexia;

· AlexiaALLAL opened on Feb 18, 2026