diff --git a/app/web_ui/src/lib/ui/info_tooltip.svelte b/app/web_ui/src/lib/ui/info_tooltip.svelte
new file mode 100644
index 00000000..30f30568
--- /dev/null
+++ b/app/web_ui/src/lib/ui/info_tooltip.svelte
@@ -0,0 +1,18 @@
+<script lang="ts">
+  export let tooltip_text: string
+</script>
+
+<button class="tooltip" data-tip={tooltip_text}>
+  <svg
+    fill="currentColor"
+    class="w-6 h-6 inline"
+    viewBox="0 0 1024 1024"
+    version="1"
+    xmlns="http://www.w3.org/2000/svg"
+    ><path
+      d="M512 717a205 205 0 1 0 0-410 205 205 0 0 0 0 410zm0 51a256 256 0 1 1 0-512 256 256 0 0 1 0 512z"
+    /><path
+      d="M485 364c7-7 16-11 27-11s20 4 27 11c8 8 11 17 11 28 0 10-3 19-11 27-7 7-16 11-27 11s-20-4-27-11c-8-8-11-17-11-27 0-11 3-20 11-28zM479 469h66v192h-66z"
+    /></svg
+  >
+</button>
diff --git a/app/web_ui/src/lib/utils/form_element.svelte b/app/web_ui/src/lib/utils/form_element.svelte
index a4163c96..fd4fc903 100644
--- a/app/web_ui/src/lib/utils/form_element.svelte
+++ b/app/web_ui/src/lib/utils/form_element.svelte
@@ -1,4 +1,6 @@
 <script lang="ts">
+  import InfoTooltip from "$lib/ui/info_tooltip.svelte"
+
   export let inputType: "input" | "textarea" | "select" = "input"
   export let id: string
   export let label: string
@@ -82,20 +84,7 @@
       >
       {#if info_description}
         <div>
-          <button class="tooltip tooltip-left" data-tip={info_description}>
-            <svg
-              fill="currentColor"
-              class="w-6 h-6 inline"
-              viewBox="0 0 1024 1024"
-              version="1"
-              xmlns="http://www.w3.org/2000/svg"
-              ><path
-                d="M512 717a205 205 0 1 0 0-410 205 205 0 0 0 0 410zm0 51a256 256 0 1 1 0-512 256 256 0 0 1 0 512z"
-              /><path
-                d="M485 364c7-7 16-11 27-11s20 4 27 11c8 8 11 17 11 28 0 10-3 19-11 27-7 7-16 11-27 11s-20-4-27-11c-8-8-11-17-11-27 0-11 3-20 11-28zM479 469h66v192h-66z"
-              /></svg
-            >
-          </button>
+          <InfoTooltip tooltip_text={info_description} />
         </div>
       {/if}
     </div>
diff --git a/app/web_ui/src/routes/(app)/run/run.svelte b/app/web_ui/src/routes/(app)/run/run.svelte
index a35603f9..aa968b5a 100644
--- a/app/web_ui/src/routes/(app)/run/run.svelte
+++ b/app/web_ui/src/routes/(app)/run/run.svelte
@@ -11,6 +11,7 @@
   import { fly } from "svelte/transition"
   import { onMount } from "svelte"
   import TagDropdown from "./tag_dropdown.svelte"
+  import InfoTooltip from "$lib/ui/info_tooltip.svelte"
 
   export let project_id: string
   export let task: Task
@@ -324,6 +325,15 @@
       delete_repair_submitting = false
     }
   }
+
+  function get_intermediate_output_title(name: string): string {
+    return (
+      {
+        reasoning: "Model Reasoning Output",
+        chain_of_thought: "Chain of Thought Output",
+      }[name] || name
+    )
+  }
 </script>
 
 <div>
@@ -345,6 +355,19 @@
         </div>
       {/if}
       <Output raw_output={run.output.output} />
+      {#if run.intermediate_outputs}
+        {#each Object.entries(run.intermediate_outputs) as [name, intermediate_output]}
+          <div
+            class="text-xs font-bold text-gray-500 mt-4 mb-1 flex flex-row items-center gap-1"
+          >
+            {get_intermediate_output_title(name)}
+            <InfoTooltip
+              tooltip_text={`This is intermediate output from the model, and not considered part of the final answer. This thinking helped formulate the final answer above. This is known as 'chain of thought', 'thinking output', or 'inference time compute'.`}
+            />
+          </div>
+          <Output raw_output={intermediate_output} />
+        {/each}
+      {/if}
       <div>
         <div class="mt-2">
           <button class="text-xs link" on:click={toggle_raw_data}
@@ -486,23 +509,9 @@
           {#each task.requirements as requirement, index}
             <div class="flex items-center">
               {requirement.name}:
-              <button
-                class="tooltip"
-                data-tip={`Requirement #${index + 1} - ${requirement.instruction || "No instruction provided"}${requirement.type === "pass_fail_critical" ? " Use 'critical' rating for responses which are never tolerable, beyond a typical failure." : ""}`}
-              >
-                <svg
-                  fill="currentColor"
-                  class="w-6 h-6 inline"
-                  viewBox="0 0 1024 1024"
-                  version="1"
-                  xmlns="http://www.w3.org/2000/svg"
-                  ><path
-                    d="M512 717a205 205 0 1 0 0-410 205 205 0 0 0 0 410zm0 51a256 256 0 1 1 0-512 256 256 0 0 1 0 512z"
-                  /><path
-                    d="M485 364c7-7 16-11 27-11s20 4 27 11c8 8 11 17 11 28 0 10-3 19-11 27-7 7-16 11-27 11s-20-4-27-11c-8-8-11-17-11-27 0-11 3-20 11-28zM479 469h66v192h-66z"
-                  /></svg
-                >
-              </button>
+              <InfoTooltip
+                tooltip_text={`Requirement #${index + 1} - ${requirement.instruction || "No instruction provided"}${requirement.type === "pass_fail_critical" ? " Use 'critical' rating for responses which are never tolerable, beyond a typical failure." : ""}`}
+              />
             </div>
             <div class="flex items-center">
               <Rating
diff --git a/libs/core/kiln_ai/adapters/fine_tune/dataset_formatter.py b/libs/core/kiln_ai/adapters/fine_tune/dataset_formatter.py
index f8c6751f..590d54f8 100644
--- a/libs/core/kiln_ai/adapters/fine_tune/dataset_formatter.py
+++ b/libs/core/kiln_ai/adapters/fine_tune/dataset_formatter.py
@@ -85,7 +85,7 @@ def build_training_data(
         include_cot
         and task_run.intermediate_outputs is not None
         and (
-            "thinking" in task_run.intermediate_outputs
+            "reasoning" in task_run.intermediate_outputs
             or "chain_of_thought" in task_run.intermediate_outputs
         )
     ):
@@ -94,7 +94,7 @@ def build_training_data(
                 "TaskRuns for training required a parent Task for building a chain of thought prompts. Train without COT, or save this TaskRun to a parent Task."
             )
         thinking = task_run.intermediate_outputs.get(
-            "thinking"
+            "reasoning"
         ) or task_run.intermediate_outputs.get("chain_of_thought")
         thinking_instructions = chain_of_thought_prompt(parent_task)
         thinking_final_answer_prompt = COT_FINAL_ANSWER_PROMPT
diff --git a/libs/core/kiln_ai/adapters/fine_tune/test_dataset_formatter.py b/libs/core/kiln_ai/adapters/fine_tune/test_dataset_formatter.py
index 6bd9dbb5..47bdb885 100644
--- a/libs/core/kiln_ai/adapters/fine_tune/test_dataset_formatter.py
+++ b/libs/core/kiln_ai/adapters/fine_tune/test_dataset_formatter.py
@@ -514,9 +514,9 @@ def test_build_training_data_with_thinking(mock_task):
     # Setup with needed fields for thinking
     mock_task_run = mock_task.runs()[0]
     assert mock_task_run.parent_task() == mock_task
-    # It should just use the thinking output if both thinking and chain_of_thought are present
+    # It should just use the reasoning output if both thinking and chain_of_thought are present
     mock_task_run.intermediate_outputs = {
-        "thinking": "thinking output",
+        "reasoning": "thinking output",
         "chain_of_thought": "cot output",
     }
     mock_task.thinking_instruction = "thinking instructions"