diff --git a/evals/eval_config.json b/evals/eval_config.json
index c51f9e79..fe50ef52 100644
--- a/evals/eval_config.json
+++ b/evals/eval_config.json
@@ -1,14 +1,15 @@
 {
     "testdata_path": "ground_truth.jsonl",
     "results_dir": "results/experiment<TIMESTAMP>",
-    "requested_metrics": ["gpt_groundedness", "gpt_relevance", "answer_length", "latency", "citation_match"],
+    "requested_metrics": ["gpt_groundedness", "gpt_relevance", "answer_length", "latency", "citations_matched"],
     "target_url": "http://127.0.0.1:8000/chat",
     "target_parameters": {
         "overrides": {
             "use_advanced_flow": true,
             "top": 3,
             "retrieval_mode": "hybrid",
-            "temperature": 0.3
+            "temperature": 0.3,
+            "seed": 42
         }
     },
     "target_response_answer_jmespath": "message.content",
diff --git a/evals/evaluate.py b/evals/evaluate.py
index 9ec2eb31..c4075479 100644
--- a/evals/evaluate.py
+++ b/evals/evaluate.py
@@ -1,17 +1,44 @@
 import argparse
 import logging
 import os
+import re
 from pathlib import Path
 from typing import Any
 
 import azure.identity
 from dotenv import load_dotenv
 from evaltools.eval.evaluate import run_evaluate_from_config
+from evaltools.eval.evaluate_metrics import register_metric
+from evaltools.eval.evaluate_metrics.base_metric import BaseMetric
 from rich.logging import RichHandler
 
 logger = logging.getLogger("ragapp")
 
 
+class CitationsMatchedMetric(BaseMetric):
+    METRIC_NAME = "citations_matched"
+
+    @classmethod
+    def evaluator_fn(cls, **kwargs):
+        def citations_overlap(*, response, ground_truth, **kwargs):
+            if response is None:
+                logger.warning("Received response of None, can't compute citation_match metric. Setting to -1.")
+                return {cls.METRIC_NAME: -1}
+            truth_citations = set(re.findall(r"\[(\d+)\]", ground_truth))
+            response_citations = set(re.findall(r"\[(\d+)\]", response))
+            # Count the percentage of citations that are present in the response
+            num_citations = len(truth_citations)
+            num_matched_citations = len(truth_citations.intersection(response_citations))
+            return {cls.METRIC_NAME: num_matched_citations / num_citations}
+
+        return citations_overlap
+
+    @classmethod
+    def get_aggregate_stats(cls, df):
+        df = df[df[cls.METRIC_NAME] != -1]
+        return {"mean": round(df[cls.METRIC_NAME].mean(), 2)}
+
+
 def get_openai_config() -> dict:
     openai_config: dict[str, Any]
     if os.environ.get("OPENAI_CHAT_HOST") == "azure":
@@ -60,6 +87,7 @@ def get_openai_config() -> dict:
 
     openai_config = get_openai_config()
 
+    register_metric(CitationsMatchedMetric)
     run_evaluate_from_config(
         working_dir=Path(__file__).parent,
         config_path="eval_config.json",
diff --git a/src/backend/fastapi_app/api_models.py b/src/backend/fastapi_app/api_models.py
index 35068a80..61027830 100644
--- a/src/backend/fastapi_app/api_models.py
+++ b/src/backend/fastapi_app/api_models.py
@@ -28,6 +28,7 @@ class ChatRequestOverrides(BaseModel):
     retrieval_mode: RetrievalMode = RetrievalMode.HYBRID
     use_advanced_flow: bool = True
     prompt_template: str | None = None
+    seed: int | None = None
 
 
 class ChatRequestContext(BaseModel):
diff --git a/src/backend/fastapi_app/rag_advanced.py b/src/backend/fastapi_app/rag_advanced.py
index b3a253fe..48ac8a09 100644
--- a/src/backend/fastapi_app/rag_advanced.py
+++ b/src/backend/fastapi_app/rag_advanced.py
@@ -35,7 +35,11 @@ def __init__(
         self.chat_token_limit = get_token_limit(chat_model, default_to_minimum=True)
 
     async def generate_search_query(
-        self, original_user_query: str, past_messages: list[ChatCompletionMessageParam], query_response_token_limit: int
+        self,
+        original_user_query: str,
+        past_messages: list[ChatCompletionMessageParam],
+        query_response_token_limit: int,
+        seed: int | None = None,
     ) -> tuple[list[ChatCompletionMessageParam], Any | str | None, list]:
         """Generate an optimized keyword search query based on the chat history and the last question"""
 
@@ -63,6 +67,7 @@ async def generate_search_query(
             n=1,
             tools=tools,
             tool_choice=tool_choice,
+            seed=seed,
         )
 
         query_text, filters = extract_search_arguments(original_user_query, chat_completion)
@@ -76,6 +81,7 @@ async def prepare_context(
             original_user_query=chat_params.original_user_query,
             past_messages=chat_params.past_messages,
             query_response_token_limit=500,
+            seed=chat_params.seed,
         )
 
         # Retrieve relevant rows from the database with the GPT optimized query
@@ -142,6 +148,7 @@ async def answer(
             max_tokens=chat_params.response_token_limit,
             n=1,
             stream=False,
+            seed=chat_params.seed,
         )
 
         return RetrievalResponse(
diff --git a/src/backend/fastapi_app/rag_base.py b/src/backend/fastapi_app/rag_base.py
index 183647e7..34fba44a 100644
--- a/src/backend/fastapi_app/rag_base.py
+++ b/src/backend/fastapi_app/rag_base.py
@@ -36,6 +36,7 @@ def get_params(self, messages: list[ChatCompletionMessageParam], overrides: Chat
         return ChatParams(
             top=overrides.top,
             temperature=overrides.temperature,
+            seed=overrides.seed,
             retrieval_mode=overrides.retrieval_mode,
             use_advanced_flow=overrides.use_advanced_flow,
             response_token_limit=response_token_limit,
diff --git a/src/backend/fastapi_app/rag_simple.py b/src/backend/fastapi_app/rag_simple.py
index 638f03ea..4bf50d1f 100644
--- a/src/backend/fastapi_app/rag_simple.py
+++ b/src/backend/fastapi_app/rag_simple.py
@@ -90,6 +90,7 @@ async def answer(
             max_tokens=chat_params.response_token_limit,
             n=1,
             stream=False,
+            seed=chat_params.seed,
         )
 
         return RetrievalResponse(
@@ -130,6 +131,7 @@ async def answer_stream(
             max_tokens=chat_params.response_token_limit,
             n=1,
             stream=True,
+            seed=chat_params.seed,
         )
 
         yield RetrievalResponseDelta(