flashrag-dev 0.1.4.dev20250306__tar.gz → 0.1.4.dev20250310__tar.gz

{flashrag_dev-0.1.4.dev20250306 → flashrag_dev-0.1.4.dev20250310}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: flashrag_dev
-Version: 0.1.4.dev20250306
+Version: 0.1.4.dev20250310
 Summary: A library for efficient Retrieval-Augmented Generation research
 Home-page: https://github.com/RUC-NLPIR/FlashRAG
 Author: Jiajie Jin, Yutao Zhu, Chenghao Zhang, Xinyu Yang, Zhicheng Dou
@@ -663,6 +663,12 @@ To facilitate easier replication of the experiments, we now provide a preprocess
 
 The index was created using the e5-base-v2 retriever on our uploaded wiki18_100w dataset, which is consistent with the index used in our experiments.
 
+## :lollipop: Awesome Work using FlashRAG
+
+- [R1-Searcher](https://github.com/SsmallSong/R1-Searcher), a method that incentivizes the search capability in LLMs via reinforcement learning
+- [ReSearch](https://github.com/Agent-RL/ReSearch), a method that learns to reason with search for LLMs via reinforcement learning
+- [AutoCoA](https://github.com/ADaM-BJTU/AutoCoA), a method that internalizes chain-of-action generation into reasoning models
+
 ## :raised_hands: Additional FAQs
 
 - [How should I set different experimental parameters?](./docs/configuration.md)

{flashrag_dev-0.1.4.dev20250306 → flashrag_dev-0.1.4.dev20250310}/README.md

@@ -646,6 +646,12 @@ To facilitate easier replication of the experiments, we now provide a preprocess
 
 The index was created using the e5-base-v2 retriever on our uploaded wiki18_100w dataset, which is consistent with the index used in our experiments.
 
+## :lollipop: Awesome Work using FlashRAG
+
+- [R1-Searcher](https://github.com/SsmallSong/R1-Searcher), a method that incentivizes the search capability in LLMs via reinforcement learning
+- [ReSearch](https://github.com/Agent-RL/ReSearch), a method that learns to reason with search for LLMs via reinforcement learning
+- [AutoCoA](https://github.com/ADaM-BJTU/AutoCoA), a method that internalizes chain-of-action generation into reasoning models
+
 ## :raised_hands: Additional FAQs
 
 - [How should I set different experimental parameters?](./docs/configuration.md)

{flashrag_dev-0.1.4.dev20250306 → flashrag_dev-0.1.4.dev20250310}/flashrag/prompt/mm_prompt.py

@@ -84,4 +84,75 @@ class GAOKAOMMPromptTemplate(MMPromptTemplate):
         content_list.append({'type': 'text', 'text': user_prompt})
         messages.append({"role": "user", "content": content_list})
         return messages
-    
+
+
+class MathVistaPromptTemplate:
+    BASE_USER_PROMPT_FREE_FORM = (
+        "You are an AI assistant designed to solve mathematical and visual reasoning problems. "
+        "Below is a question that requires you to analyze the given information and provide a detailed answer.\n\n"
+        "### Reference Information:\n{reference}\n\n"
+        "### Question:\n{question}\n\n"
+        "### Instructions:\n"
+        "Only give me the final answer.\n\n"
+        "### Answer:\n"
+    )
+
+    BASE_USER_PROMPT_MULTI_CHOICE = (
+        "You are an AI assistant designed to solve mathematical and visual reasoning problems. "
+        "Below is a multiple-choice question that requires you to analyze the given information and select the correct choice.\n\n"
+        "### Reference Information:\n{reference}\n\n"
+        "### Question:\n{question}\n\n"
+        "### Instructions:\n"
+        "Only give me the correct option letter, e.g., A, B, C, D.\n\n"
+        "### Answer:\n"
+    )
+
+    def __init__(self, config, system_prompt=None, user_prompt_free_form=None, user_prompt_multi_choice=None):
+        self.config = config
+        self.system_prompt = system_prompt
+        self.user_prompt_free_form = user_prompt_free_form if user_prompt_free_form is not None else self.BASE_USER_PROMPT_FREE_FORM
+        self.user_prompt_multi_choice = user_prompt_multi_choice if user_prompt_multi_choice is not None else self.BASE_USER_PROMPT_MULTI_CHOICE
+
+    def get_string(self, item, use_retrieval_image=False, use_retrieval_text=False, use_question_image=False):
+        question = item.question if item.question is not None else item.text
+        question_image = item.image
+        question_type = item.question_type  # 'free_form' or 'multi_choice'
+
+        messages = []
+        if self.system_prompt is not None:
+            messages.append({"role": "system", "content": self.system_prompt})
+
+        content_list = []
+        reference_str = ""
+
+        # Add retrieval results as reference (if available)
+        try:
+            retrieval_result = item.retrieval_result
+        except:
+            retrieval_result = []
+            
+        for idx, item in enumerate(retrieval_result):
+            if 'image' not in item:
+                reference_str += f'Example {idx + 1}: {item["contents"]}\n'
+            else:
+                content_list.append({'type': 'image', 'image': item['image']})
+                reference_str += f'Example {idx + 1}: {item["text"]}\n'
+
+        # Add question image (if available)
+        if question_image:
+            content_list.append({'type': 'image', 'image': question_image})
+
+        # Format the user prompt based on question type
+        if question_type == 'free_form':
+            user_prompt = self.user_prompt_free_form.format(question=question, reference=reference_str)
+        elif question_type == 'multi_choice':
+            user_prompt = self.user_prompt_multi_choice.format(question=question, reference=reference_str)
+        else:
+            raise ValueError(f"Unsupported question type: {question_type}")
+
+        # Add the formatted user prompt
+        content_list.append({'type': 'text', 'text': user_prompt})
+        messages.append({"role": "user", "content": content_list})
+
+        return messages
+    

{flashrag_dev-0.1.4.dev20250306 → flashrag_dev-0.1.4.dev20250310}/flashrag/retriever/retriever.py

@@ -324,7 +324,8 @@ class BM25Retriever(BaseTextRetriever):
             results, scores = self.searcher.retrieve(query_tokens, k=num)
         else:
             assert False, "Invalid bm25 backend!"
-
+        results = results.tolist() if isinstance(results,np.ndarray) else results
+        scores = scores.tolist() if isinstance(scores,np.ndarray) else scores
         if return_score:
             return results, scores
         else:

flashrag_dev-0.1.4.dev20250310/flashrag/version.py

@@ -0,0 +1 @@
+__version__ = "0.1.4dev20250310"

{flashrag_dev-0.1.4.dev20250306 → flashrag_dev-0.1.4.dev20250310}/flashrag_dev.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: flashrag-dev
-Version: 0.1.4.dev20250306
+Version: 0.1.4.dev20250310
 Summary: A library for efficient Retrieval-Augmented Generation research
 Home-page: https://github.com/RUC-NLPIR/FlashRAG
 Author: Jiajie Jin, Yutao Zhu, Chenghao Zhang, Xinyu Yang, Zhicheng Dou
@@ -663,6 +663,12 @@ To facilitate easier replication of the experiments, we now provide a preprocess
 
 The index was created using the e5-base-v2 retriever on our uploaded wiki18_100w dataset, which is consistent with the index used in our experiments.
 
+## :lollipop: Awesome Work using FlashRAG
+
+- [R1-Searcher](https://github.com/SsmallSong/R1-Searcher), a method that incentivizes the search capability in LLMs via reinforcement learning
+- [ReSearch](https://github.com/Agent-RL/ReSearch), a method that learns to reason with search for LLMs via reinforcement learning
+- [AutoCoA](https://github.com/ADaM-BJTU/AutoCoA), a method that internalizes chain-of-action generation into reasoning models
+
 ## :raised_hands: Additional FAQs
 
 - [How should I set different experimental parameters?](./docs/configuration.md)

flashrag_dev-0.1.4.dev20250306/flashrag/version.py

@@ -1 +0,0 @@
-__version__ = "0.1.4dev20250306"