From 956f952063093ef85644e55d5b18a37c8c45f918 Mon Sep 17 00:00:00 2001 From: BM-K Date: Thu, 15 Jan 2026 06:28:59 +0000 Subject: [PATCH 1/2] model: add pixie_models --- .../model_implementations/pixie_models.py | 58 +++++++++++++++++++ 1 file changed, 58 insertions(+) create mode 100644 mteb/models/model_implementations/pixie_models.py diff --git a/mteb/models/model_implementations/pixie_models.py b/mteb/models/model_implementations/pixie_models.py new file mode 100644 index 0000000000..c67c04bd4d --- /dev/null +++ b/mteb/models/model_implementations/pixie_models.py @@ -0,0 +1,58 @@ +from mteb.models.model_meta import ( + ModelMeta, + ScoringFunction, +) +from mteb.models.sentence_transformer_wrapper import sentence_transformers_loader + +from mteb.models.model_implementations.arctic_models import ( + ARCTIC_V2_CITATION, + LANGUAGES_V2_0, + arctic_v2_training_datasets, +) + +PIXIE_RUNE_V1_CITATION = """@misc{TelePIX-PIXIE-Rune-v1.0, + title = {PIXIE-Rune-v1.0}, + author = {TelePIX AI Research Team and Bongmin Kim}, + year = {2026}, + howpublished = {Hugging Face model card}, + url = {https://huggingface.co/telepix/PIXIE-Rune-v1.0} +}""" + +PIXIE_RUNE_V1_PROMPTS = { + "query": "query: ", + "document": "", +} + +# it is further fine-tuned on TelePIX proprietary IR data (not public). +pixie_rune_v1_training_datasets = set(arctic_v2_training_datasets) | { + "TelePIX-Proprietary-IR-Triplets", +} + +pixie_rune_v1_0 = ModelMeta( + loader=sentence_transformers_loader, + loader_kwargs={ + "model_prompts": PIXIE_RUNE_V1_PROMPTS, + }, + name="telepix/PIXIE-Rune-v1.0", + model_type=["dense"], + revision="b2486496da71191626666a88f9bfec844933a134", + release_date="2026-01-15", + languages=LANGUAGES_V2_0, + open_weights=True, + framework=["Sentence Transformers", "PyTorch", "safetensors"], + n_parameters=567754752, + memory_usage_mb=2166, + max_tokens=6144, + embed_dim=1024, + license="apache-2.0", + reference="https://huggingface.co/telepix/PIXIE-Rune-v1.0", + similarity_fn_name=ScoringFunction.COSINE, + use_instructions=True, + adapted_from="Snowflake/snowflake-arctic-embed-l-v2.0", + superseded_by=None, + public_training_code=None, + public_training_data=None, + training_datasets=pixie_rune_v1_training_datasets, + citation=PIXIE_RUNE_V1_CITATION + "\n\n" + ARCTIC_V2_CITATION, +) + From c042e27a63931e63331c0713edfad8c90e63f3b3 Mon Sep 17 00:00:00 2001 From: BM-K Date: Thu, 15 Jan 2026 09:12:10 +0000 Subject: [PATCH 2/2] Apply lint formatting --- mteb/models/model_implementations/pixie_models.py | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-) diff --git a/mteb/models/model_implementations/pixie_models.py b/mteb/models/model_implementations/pixie_models.py index c67c04bd4d..8c8a058995 100644 --- a/mteb/models/model_implementations/pixie_models.py +++ b/mteb/models/model_implementations/pixie_models.py @@ -1,14 +1,13 @@ -from mteb.models.model_meta import ( - ModelMeta, - ScoringFunction, -) -from mteb.models.sentence_transformer_wrapper import sentence_transformers_loader - from mteb.models.model_implementations.arctic_models import ( ARCTIC_V2_CITATION, LANGUAGES_V2_0, arctic_v2_training_datasets, ) +from mteb.models.model_meta import ( + ModelMeta, + ScoringFunction, +) +from mteb.models.sentence_transformer_wrapper import sentence_transformers_loader PIXIE_RUNE_V1_CITATION = """@misc{TelePIX-PIXIE-Rune-v1.0, title = {PIXIE-Rune-v1.0}, @@ -55,4 +54,3 @@ training_datasets=pixie_rune_v1_training_datasets, citation=PIXIE_RUNE_V1_CITATION + "\n\n" + ARCTIC_V2_CITATION, ) -