fix: change passage prompt to document

mteb/encoder_interface.py

@@ -14,7 +14,7 @@
 
 class PromptType(str, Enum):
     query = "query"
-    passage = "passage"
+    document = "document"
 
 
 @runtime_checkable

mteb/evaluation/evaluators/Image/Any2AnyRetrievalEvaluator.py

@@ -195,7 +195,7 @@ def search(
                 sub_corpus_embeddings = self.model.get_text_embeddings(
                     texts=corpus_texts,
                     task_name=task_name,
-                    prompt_type=PromptType.passage,
+                    prompt_type=PromptType.document,
                     **self.encode_kwargs,
                 )
             else:
@@ -213,7 +213,7 @@ def search(
                     sub_corpus_embeddings = self.model.get_image_embeddings(
                         images=corpus_image_dataloader,
                         task_name=task_name,
-                        prompt_type=PromptType.passage,
+                        prompt_type=PromptType.document,
                         **self.encode_kwargs,
                     )
                 elif corpus_modality == "image,text":
@@ -222,7 +222,7 @@ def search(
                         texts=corpus_texts,
                         images=corpus_image_dataloader,
                         task_name=task_name,
-                        prompt_type=PromptType.passage,
+                        prompt_type=PromptType.document,
                         **self.encode_kwargs,
                     )
                 else:

mteb/evaluation/evaluators/RerankingEvaluator.py

@@ -180,7 +180,7 @@ def _encode_candidates_batched(
             all_docs,
             model,
             task_name=self.task_name,
-            prompt_type=PromptType.passage,
+            prompt_type=PromptType.document,
             **self.encode_kwargs,
         )
 
@@ -245,7 +245,7 @@ def _encode_candidates_individual(
                 model.encode(
                     docs,
                     task_name=self.task_name,
-                    prompt_type=PromptType.passage,
+                    prompt_type=PromptType.document,
                     **self.encode_kwargs,
                 )
             )
@@ -293,7 +293,7 @@ def _encode_candidates_miracl_batched(self, all_query_embs, model: Encoder):
             model.encode(
                 all_docs,
                 task_name=self.task_name,
-                prompt_type=PromptType.passage,
+                prompt_type=PromptType.document,
                 **self.encode_kwargs,
             )
         )
@@ -345,7 +345,7 @@ def _encode_candidates_miracl_individual(self, model: Encoder):
                     model.encode(
                         docs,
                         task_name=self.task_name,
-                        prompt_type=PromptType.passage,
+                        prompt_type=PromptType.document,
                         **self.encode_kwargs,
                     )
                 )

mteb/evaluation/evaluators/RetrievalEvaluator.py

@@ -159,7 +159,7 @@ def search(
                 sub_corpus_embeddings = self.model.encode(
                     corpus[corpus_start_idx:corpus_end_idx],  # type: ignore
                     task_name=task_name,
-                    prompt_type=PromptType.passage,
+                    prompt_type=PromptType.document,
                     request_qid=request_qid,
                     **self.encode_kwargs,
                 )
@@ -385,7 +385,7 @@ def encode_corpus(
         corpus: list[dict[str, str]],
         task_name: str,
         batch_size: int,
-        prompt_type: PromptType = PromptType.passage,
+        prompt_type: PromptType = PromptType.document,
         request_qid: str | None = None,
         **kwargs,
     ):
@@ -416,7 +416,7 @@ def encode(
         prompt_type: PromptType | None = None,
         **kwargs,
     ):
-        if prompt_type and prompt_type == PromptType.passage:
+        if prompt_type and prompt_type == PromptType.document:
             return self.encode_corpus(
                 sentences, task_name, prompt_type=prompt_type, **kwargs
             )

mteb/models/cadet_models.py

@@ -35,7 +35,7 @@
         revision="8056d118be37a566f20972a5f35cda815f6bc47e",
         model_prompts={
             "query": "query: ",
-            "passage": "passage: ",
+            "document": "passage: ",
         },
     ),
     name="manveertamber/cadet-embed-base-v1",

mteb/models/cohere_models.py

@@ -211,7 +211,7 @@ def encode(
     "MultilabelClassification": "classification",
     "Clustering": "clustering",
     PromptType.query.value: "search_query",
-    PromptType.passage.value: "search_document",
+    PromptType.document.value: "search_document",
 }
 
 cohere_mult_3 = ModelMeta(

mteb/models/e5_models.py

@@ -110,7 +110,7 @@
 
 model_prompts = {
     PromptType.query.value: "query: ",
-    PromptType.passage.value: "passage: ",
+    PromptType.document.value: "passage: ",
 }
 
 E5_TRAINING_DATA = {

mteb/models/gme_v_models.py

@@ -182,7 +182,7 @@ def encode_corpus(self, corpus: list[dict[str, str]], **kwargs):
                 else doc["text"].strip()
                 for doc in corpus
             ]
-        embeddings = self.encode(sentences, prompt_type=PromptType.passage**kwargs)
+        embeddings = self.encode(sentences, prompt_type=PromptType.document**kwargs)
         return embeddings
 
     def get_image_embeddings(self, images: list[Image.Image] | DataLoader, **kwargs):
@@ -210,7 +210,7 @@ def get_fused_embeddings(
         instruction=None,
         **kwargs: Any,
     ):
-        if prompt_type == PromptType.passage:
+        if prompt_type == PromptType.document:
             instruction = None
         elif instruction is None:
             instruction = self.get_instruction(task_name, prompt_type)

mteb/models/google_models.py

@@ -39,7 +39,7 @@
     "Clustering": "CLUSTERING",
     "STS": "SEMANTIC_SIMILARITY",
     PromptType.query.value: "RETRIEVAL_QUERY",
-    PromptType.passage.value: "RETRIEVAL_DOCUMENT",
+    PromptType.document.value: "RETRIEVAL_DOCUMENT",
 }
 
 GECKO_TRAINING_DATA = {

mteb/models/instruct_wrapper.py

@@ -147,10 +147,13 @@ def encode(
         )
 
         # to passage prompts won't be applied to passages
-        if not self.apply_instruction_to_passages and prompt_type == PromptType.passage:
+        if (
+            not self.apply_instruction_to_passages
+            and prompt_type == PromptType.document
+        ):
             instruction = None
             logger.info(
-                f"No instruction used, because prompt type = {prompt_type.passage}"
+                f"No instruction used, because prompt type = {prompt_type.document}"
             )
 
         if instruction:

mteb/models/jasper_models.py

@@ -44,7 +44,7 @@ def encode(
         instruction = self.get_task_instruction(task_name, prompt_type)
 
         # to passage prompts won't be applied to passages
-        if prompt_type == PromptType.passage and task.metadata.category == "s2p":
+        if prompt_type == PromptType.document and task.metadata.category == "s2p":
             instruction = None
 
         embeddings = self.model.encode(

mteb/models/jina_models.py

@@ -269,8 +269,8 @@ def _resolve_task_parameters(
         # Determine prompt name parameter
         if jina_task_name and "query" in jina_task_name:
             prompt_name_param = "query"
-        elif jina_task_name and "passage" in jina_task_name:
-            prompt_name_param = "passage"
+        elif jina_task_name and "document" in jina_task_name:
+            prompt_name_param = "document"
         else:
             prompt_name_param = "query"  # default fallback
 
@@ -549,7 +549,7 @@ def get_programming_task_override(
         trust_remote_code=True,
         model_prompts={
             "Retrieval-query": "retrieval.query",
-            "Retrieval-passage": "retrieval.passage",
+            "Retrieval-document": "retrieval.passage",
             "STS": "text-matching",
             "DocumentUnderstanding": "retrieval.query",
         },
@@ -584,7 +584,7 @@ def get_programming_task_override(
         trust_remote_code=True,
         model_prompts={
             "Retrieval-query": "retrieval.query",
-            "Retrieval-passage": "retrieval.passage",
+            "Retrieval-document": "retrieval.passage",
             "Clustering": "separation",
             "Classification": "classification",
             "STS": "text-matching",