2.10.12
2.10.12 (2026-03-14)
Documentation
- docs: remove mieb and mmteb contribution docs (#4227)
I don't think we maintain these anymore. I think they are fine to remove (823236c)
- docs: fix docs paths (#4224)
fix docs paths (0d07d33)
- docs: fix naming on contributing docs (
6787a17)
Fix
-
fix: metadata getting computed for existing MTEB model (#4231)
-
Fix behaviour while getting metadata of existing MTEB model
-
Added basic metadata in overwrite
-
Updated CrossEncoderWrapper with same changes (
973a5a1)
Unknown
- Model: Add new model revision of Querit/Querit (#4215)
New model revision (f913ed8)
- Fix zeroentropy/zembed-1 metadata (#4233)
Fix zeroentropy/zembed-1 metadata (revision, release_date, max_tokens)
The metadata added in #4202 had incorrect values for three fields:
- revision: pointed to wrong HuggingFace commit
- release_date: was "2025-09-16", should be "2026-03-02"
- max_tokens: was 40960, should be 32768 (
3cd67fd)
-
Add Zeroentropy models (#4228)
-
Add Zeroentropy models
-
correct metadata
-
Correct loader_kwargs for rerankers (
791a185) -
model: nvidia/llama-nemotron-embed-vl-1b-v2 for ViDoRe (#4192)
-
Adds nvidia/llama-nemotron-embed-vl-1b-v2 model
-
Update mteb/models/model_implementations/nvidia_nemotron_vl_models.py
Co-authored-by: Roman Solomatin <samoed.roman@gmail.com>
- Update mteb/models/model_implementations/nvidia_nemotron_vl_models.py
Co-authored-by: Roman Solomatin <samoed.roman@gmail.com>
- Update mteb/models/model_implementations/nvidia_nemotron_vl_models.py
Co-authored-by: Roman Solomatin <samoed.roman@gmail.com>
- Update mteb/models/model_implementations/nvidia_nemotron_vl_models.py
Co-authored-by: Roman Solomatin <samoed.roman@gmail.com>
- Update mteb/models/model_implementations/nvidia_nemotron_vl_models.py
Co-authored-by: Roman Solomatin <samoed.roman@gmail.com>
- Update mteb/models/model_implementations/nvidia_nemotron_vl_models.py
Co-authored-by: Roman Solomatin <samoed.roman@gmail.com>
-
Fixing tests and linting issues
-
Update mteb/models/model_implementations/nvidia_nemotron_vl_models.py
Co-authored-by: Roman Solomatin <samoed.roman@gmail.com>
-
Nemotron Embed VL 1B: Setting the number of tiles an image can be split
-
Fixing lint issue
-
Update mteb/models/model_implementations/nvidia_nemotron_vl_models.py
-
Disabling image modality by default
-
Update mteb/models/model_implementations/nvidia_nemotron_vl_models.py
Co-authored-by: Roman Solomatin <samoed.roman@gmail.com>
- Setting default modality of Nemotron Embed VL 1B as image + text (when available)
Co-authored-by: Roman Solomatin <samoed.roman@gmail.com> (2a8c2d3)