diff --git a/tests/unit/vertexai/genai/test_evals.py b/tests/unit/vertexai/genai/test_evals.py
index 2c6442818d..d72b48a2fc 100644
--- a/tests/unit/vertexai/genai/test_evals.py
+++ b/tests/unit/vertexai/genai/test_evals.py
@@ -3749,7 +3749,7 @@ def test_eval_case_to_agent_data_events_only(self):
         assert agent_data.agent_config is None
         assert agent_data.events.event[0].parts[0].text == "intermediate event"
 
-    def test_eval_case_to_agent_data_empty_events(self):
+    def test_eval_case_to_agent_data_empty_event_content(self):
         intermediate_events = [
             vertexai_genai_types.Event(
                 event_id="event1",
@@ -3774,8 +3774,86 @@ def test_eval_case_to_agent_data_empty_events(self):
         )
 
         assert agent_data.agent_config is None
-        assert agent_data.events is None
-        assert not agent_data.events_text
+        assert not agent_data.events.event
+
+    def test_eval_case_to_agent_data_empty_intermediate_events_list(self):
+        agent_info = vertexai_genai_types.evals.AgentInfo(
+            name="agent1",
+            instruction="instruction1",
+            tool_declarations=[],
+        )
+
+        eval_case = vertexai_genai_types.EvalCase(
+            prompt=genai_types.Content(parts=[genai_types.Part(text="Hello")]),
+            responses=[
+                vertexai_genai_types.ResponseCandidate(
+                    response=genai_types.Content(parts=[genai_types.Part(text="Hi")])
+                )
+            ],
+            agent_info=agent_info,
+        )
+
+        agent_data = (
+            _evals_metric_handlers.PredefinedMetricHandler._eval_case_to_agent_data(
+                eval_case
+            )
+        )
+
+        assert not agent_data.events.event
+
+    def test_eval_case_to_agent_data_agent_info_empty_tools(self):
+        agent_info = vertexai_genai_types.evals.AgentInfo(
+            name="agent1",
+            instruction="instruction1",
+            tool_declarations=[],
+        )
+        eval_case = vertexai_genai_types.EvalCase(
+            prompt=genai_types.Content(parts=[genai_types.Part(text="Hello")]),
+            responses=[
+                vertexai_genai_types.ResponseCandidate(
+                    response=genai_types.Content(parts=[genai_types.Part(text="Hi")])
+                )
+            ],
+            agent_info=agent_info,
+            intermediate_events=None,
+        )
+
+        agent_data = (
+            _evals_metric_handlers.PredefinedMetricHandler._eval_case_to_agent_data(
+                eval_case
+            )
+        )
+
+        assert agent_data.agent_config.developer_instruction.text == "instruction1"
+        assert not agent_data.agent_config.tools.tool
+
+    def test_eval_case_to_agent_data_agent_info_empty(self):
+        intermediate_events = [
+            vertexai_genai_types.Event(
+                event_id="event1",
+                content=genai_types.Content(
+                    parts=[genai_types.Part(text="intermediate event")]
+                ),
+            )
+        ]
+        eval_case = vertexai_genai_types.EvalCase(
+            prompt=genai_types.Content(parts=[genai_types.Part(text="Hello")]),
+            responses=[
+                vertexai_genai_types.ResponseCandidate(
+                    response=genai_types.Content(parts=[genai_types.Part(text="Hi")])
+                )
+            ],
+            agent_info=None,
+            intermediate_events=intermediate_events,
+        )
+
+        agent_data = (
+            _evals_metric_handlers.PredefinedMetricHandler._eval_case_to_agent_data(
+                eval_case
+            )
+        )
+
+        assert agent_data.agent_config is None
 
 
 @pytest.mark.usefixtures("google_auth_mock")
diff --git a/vertexai/_genai/_evals_data_converters.py b/vertexai/_genai/_evals_data_converters.py
index 340bf72965..337abaaae8 100644
--- a/vertexai/_genai/_evals_data_converters.py
+++ b/vertexai/_genai/_evals_data_converters.py
@@ -366,10 +366,6 @@ def convert(self, raw_data: list[dict[str, Any]]) -> types.EvaluationDataset:
 
             intermediate_events: Optional[list[types.Event]] = None
             if intermediate_events_data:
-                logger.warning(
-                    "intermediate_events attribute is experimental and may change in "
-                    "future versions."
-                )
                 if isinstance(intermediate_events_data, list):
                     intermediate_events = []
                     for event in intermediate_events_data:
diff --git a/vertexai/_genai/_evals_metric_handlers.py b/vertexai/_genai/_evals_metric_handlers.py
index acbeda2afd..322d3aff71 100644
--- a/vertexai/_genai/_evals_metric_handlers.py
+++ b/vertexai/_genai/_evals_metric_handlers.py
@@ -854,8 +854,9 @@ def _eval_case_to_agent_data(
             return None
         tools = None
         developer_instruction = None
-        events = None
         agent_config = None
+        tool_declarations = []
+        event_contents = []
 
         if eval_case.agent_info:
             agent_info = eval_case.agent_info
@@ -865,7 +866,8 @@ def _eval_case_to_agent_data(
                 )
             if agent_info.tool_declarations:
                 tool_declarations = agent_info.tool_declarations
-                tools = types.evals.Tools(tool=tool_declarations)
+            tools = types.evals.Tools(tool=tool_declarations)
+
             if tools or developer_instruction:
                 agent_config = types.evals.AgentConfig(
                     tools=tools,
@@ -878,19 +880,12 @@ def _eval_case_to_agent_data(
                 for event in eval_case.intermediate_events
                 if event.content
             ]
-            if event_contents:
-                events = types.evals.Events(event=event_contents)
+        events = types.evals.Events(event=event_contents)
 
-        if events:
-            return types.evals.AgentData(
-                agent_config=agent_config,
-                events=events,
-            )
-        else:
-            return types.evals.AgentData(
-                agent_config=agent_config,
-                events_text="",
-            )
+        return types.evals.AgentData(
+            agent_config=agent_config,
+            events=events,
+        )
 
     def _build_request_payload(
         self, eval_case: types.EvalCase, response_index: int
diff --git a/vertexai/_genai/_evals_visualization.py b/vertexai/_genai/_evals_visualization.py
index 0436d893ae..f5e19d430a 100644
--- a/vertexai/_genai/_evals_visualization.py
+++ b/vertexai/_genai/_evals_visualization.py
@@ -280,7 +280,7 @@ def _get_evaluation_html(eval_result_json: str) -> str:
 
             // If we have agent info, render as trace
             if(agentInfo) {{
-                let traceHtml = `<div class="trace-event-row"><div class="name"><span class="icon">🏃</span>agent_run</div></div>`;
+                let traceHtml = `<div class="trace-event-row"><div class="name"><span class="icon">🤖</span>agent_run</div></div>`;
                 eventsArray.forEach(event => {{
                     if (event.content && event.content.parts && event.content.parts.length > 0) {{
                         event.content.parts.forEach(part => {{