Mteb

Ci

* ci: fix outdated CI for pandas (1029) ([`039c976`](https://github.com/embeddings-benchmark/mteb/commit/039c9765d6918ce375fea7aaa5e393ade12dac9d))

Fix

* fix: parallel bitext tasks scores (1028)

* fix: Added imenes fix

* fix: found and fixes two other mistakes for MLSUMClustering (one caused by python 3.8 and the other in the downsampling function)

* fix: parallel bitext tasks scores

Due to the implementation of parallel bitext tasks it computes all combinations even though only some are specified by eval_langs. This leads to a minor computational overhead when evaluating and most notably an error when constructing MTEBResults.

This fix does the following:
- it ensures that only desired language pairs are being evaluated
- it simplified bitext evaluation (there is no longer two cases)
- it additionally only encodes the sentence pairs required for the specific selected pairs (make it faster when we only want to consider a subset of languages)

I tested it using:
- A monolingual bitext task
- A multilingual parallel bitext task
- A multilingal non-parallel bitext task

All pass and reproduce the same results (cheers!)

* Update mteb/abstasks/AbsTaskBitextMining.py

Co-authored-by: Isaac Chung <chungisaac1217gmail.com>

---------

Co-authored-by: Isaac Chung <chungisaac1217gmail.com> ([`3e5ba61`](https://github.com/embeddings-benchmark/mteb/commit/3e5ba61de0d8b3081f394f1c618937769417d0b8))

Unknown

* Merge branch &39;main&39; of https://github.com/embeddings-benchmark/mteb ([`e02fe6e`](https://github.com/embeddings-benchmark/mteb/commit/e02fe6e105391d57863b8cf4f3feba006b7213a4))

Fix

* fix: The Swahili Classification Task (998)

* swahili_news_classfication

* update_two

* Update Swahili addin source and refeence

* Update SwahiliNewsClassification.py

* Update SwahiliNewsClassification.py

* addition info about group

* remove unnecessary line ([`7e4ba06`](https://github.com/embeddings-benchmark/mteb/commit/7e4ba0634feb41b27c9e17b3adbe0e1df914af4b))

Documentation

* docs: Update figure (1027)

* Updated figure colors

* Updated figure with current numbers

* Added reasoning as retrieval to figure ([`0acf259`](https://github.com/embeddings-benchmark/mteb/commit/0acf2590b9b6059fc5dbe66c31b1fc5f313efdd6))

* docs: Added result tables (996)

* Added utility function for downloading models results

* Added result tables ([`f809cad`](https://github.com/embeddings-benchmark/mteb/commit/f809cadcbaa259f8bbfc5d3fe98de217023883f1))

* docs: Add Mariya&39;s points (1016)

* Create 1004.jsonl (1011)

add points for working on introduction

* Create 784.jsonl (1012)

add points for working on Appendix B

* rename files

* fix: Fixed subset warning for later versions of polars

---------

Co-authored-by: Mariya Hendriksen <35101262+mariyahendriksenusers.noreply.github.com> ([`4e1f43b`](https://github.com/embeddings-benchmark/mteb/commit/4e1f43b1d56843fe36637068d64e5a09e63964fd))

* docs: Update points.md (1010)

Added an author, fixed alignment issues and some typos
Co-authored-by: dokato <raymon92gmail.com> ([`1d1f7aa`](https://github.com/embeddings-benchmark/mteb/commit/1d1f7aadc71b89444837ea59bd142107789f648a))

Fix

* fix: Fix broken clustering tasks (1026)

* fix: Added imenes fix

* fix: found and fixes two other mistakes for MLSUMClustering (one caused by python 3.8 and the other in the downsampling function) ([`ca9e81d`](https://github.com/embeddings-benchmark/mteb/commit/ca9e81de8e950eef823e31e6799fe9e3fa5767dd))

Unknown

* LLM2Vec multi-gpu bug fix (1020)

* map device to device_map

* lint ([`3a64ed9`](https://github.com/embeddings-benchmark/mteb/commit/3a64ed9ffe87a8673b6c5ebe05bb0c095a2d1698))

* Update points table ([`b3a58d2`](https://github.com/embeddings-benchmark/mteb/commit/b3a58d2ab40cfee130aba2784ba9a067a929ad4f))

* add SFR embed model (1002)

* add sfr model
* add instruction to encode ([`1e25fe0`](https://github.com/embeddings-benchmark/mteb/commit/1e25fe04cd024b3032c80c1479f07f661a788d3f))

Fix

* fix: Add more russian models (1000)

* add models meta

* add models

* fix models versions

* fix models versions

* add models

* add points

* lint

* add to points table ([`d97310d`](https://github.com/embeddings-benchmark/mteb/commit/d97310d6bc090d4e7bb0fde93e52b9241044bc7e))

Unknown

* Update points table ([`17262fe`](https://github.com/embeddings-benchmark/mteb/commit/17262fe895c3f1cc3b6119c42754294d2a3a1851))

Fix

* fix: Added utility function for downloading and loading models results (995)

Added utility function for downloading models results ([`abef8dc`](https://github.com/embeddings-benchmark/mteb/commit/abef8dc2366d56c790a99ed205b97fc81d0894c2))

Unknown

* Update points table ([`17308b3`](https://github.com/embeddings-benchmark/mteb/commit/17308b3d0634a2baae373586f30f34f83bbee7ec))

* [WIP] RAR-b second PR & task instruction for IR tasks in general (972)

* Reval with instruction ARC-C test

* fix formatting issue

* add default instructions to lookup

* add points for both PRs

* add points for both PRs & info ([`4500e0d`](https://github.com/embeddings-benchmark/mteb/commit/4500e0d1157f1bc291dade1f558480fe620c5b9c))

* Update tasks table ([`be5bf69`](https://github.com/embeddings-benchmark/mteb/commit/be5bf694479ed26d30ba554363c80ad4033fb84c))