Added Ministral 3 tip to enable reasoning.

index.html

@@ -86,7 +86,7 @@
 			<li class="⚙️">Top_nSigma 1.25 (this sampler enable better higher temperature and creativity, without the drawbacks)</li>
 			<li class="⚙️">Since I'm usually using LM Studio as a backend, I've not played much with the excellent XTC, DRY and nSignma samplers.</li>
 
-		<li class="🤖"><emo>🍺</emo> Most universal <b>context &amp; instruct templates</b>: ChatML, or ChatML <emo>💭</emo>Reasoning </li>
+		<li class="🤖"><emo>🍺</emo> Most universal <b>context &amp; instruct templates</b>: ChatML, or ChatML Reasoning </li>
 			<li class="⚙️">Very old models can also be used with the Alpaca template.</li>
 			<li class="⚙️">For most recent models, connecting to your backend via "Chat Completion API" removes the need to select a template.</li>
 
@@ -293,7 +293,7 @@
 	<li class="🏢" id="🏢Xiaomi"><i>Xiaomi</i><flag>🇨🇳</flag></li>
 
 		<li class="🤖" id="🤖MiMo 2 Flash"><b>MiMo 2 Flash</b></li>
-			<li class="⚖️ 👎"><b>Clueless at Roleplay</b> <emo>😵</emo> The model was clearly not designed for roleplay. At least with the Q2_K_XL version I tested locally, responses were unatural, prone to looping, and emotionaly flat.</li>
+			<li class="⚖️ 👎"><b>Clueless at Roleplay</b> <emo>😵</emo> The model was clearly not designed with roleplay in mind. At least with the Q2_K_XL version I tested locally, responses were unatural, prone to looping, and emotionaly flat.</li>
 			<li class="⚙️">Temperature 0.8</li>
 			<li class="⚙️">Top_P 0.95</li>
 			<li class="⚙️">You'll have to use Chat Completion API to connect</li>
@@ -325,25 +325,31 @@
 			<li class="▶️▶️ ⚙️">Temperature 0.15 or 0.1</li>
 			<li class="▶️▶️ ⚙️">Top_P 1.0</li>
 		<li class="▶️"><emo>💭</emo> Reasoning usage</li>
-			<li class="▶️▶️ ⚙️">Temperature 0.17</li>
+			<li class="▶️▶️ ⚙️">Temperature 0.7</li>
 			<li class="▶️▶️ ⚙️">Top_P 0.95</li>
 			<li class="🍺"><emo>💭</emo> Reasoning Formatting: [THINK] [/THINK]</li>
-			<li class="🍺"><emo>🍺</emo> Not sure yet how to trigger reasonning in Silly Tavern</li>
+			<li class="🍺"><emo>🍺</emo> To trigger reasonning in SillyTavern: 'Start replies with' <em>[THINK]</em> and add the following to your prompt:<br>
+							<pre style="white-space: inherit;">&lt;s&gt;[SYSTEM_PROMPT]# HOW YOU SHOULD THINK AND ANSWER
+
+First draft your thinking process (inner monologue) until you arrive at a response. Format your response using Markdown, and use LaTeX for any mathematical equations. Write both your thoughts and the response in the same language as the input.
+
+Your thinking process must follow the template below:[THINK]Your thoughts or/and draft, like working through an exercise on scratch paper. Be as casual and as long as you want until you are confident to generate the response to the user.[/THINK]Here, provide a self-contained response.[/SYSTEM_PROMPT][INST]What is 1+1?[/INST]2&lt;s&gt;[INST]What is 2+2?[/INST]</pre></li>
 
-		<li class="🍺"><emo>🍺</emo> If using Chat Connection API, set "Prompt Post-Processing" to Strict</li>
+		<li class="🍺"><emo>🍺</emo> If using Chat Completion API, set "Prompt Post-Processing" to Strict</li>
 			<li class="🍺"><emo>🍺</emo> Ramble way too much? Use <a href="https://huggingface.co/sleepdeprived3/Mistral-V7-Tekken-Concise">Mistral-V7-Tekken-Concise prompt</a></li>
-			<li class="🍺"><emo>🍺</emo> Guide to selecting the correct Mistral template</li>
+			<li class="🍺"><emo>🍺</emo> Guide to selecting the correct Mistral template 👇</li>
 			<li class="💥"><emo>🔞</emo><emo>💥</emo> No need to use a jailbreak prompt, the model is already extremely horny by default!</li>
 
 		<li class="🤖" id="🤖Mistral Large"><b>Mistral Large</b></li>
 			<li class="⚙️">Temperature 0.7</li>
 			<li class="⚙️">Do not use quantize KV cache</li>
-			<li class="🍺"><emo>🍺</emo> Guide to selecting the correct Mistral template</li>
+			<li class="🍺"><emo>🍺</emo> Guide to selecting the correct Mistral template 👇</li>
 
 		<li class="🤖" id="🤖Mistral Small 3.x"><b>Mistral Small 3.x</b></li>
+			<li class="⚖️ 👎"><b>Verbose</b> <emo>🗣️</emo> Models based on Mistral 3.1 and 3.2 tends to write walls of text.</li>
 			<li class="⚙️">Temperature 0.15</li>
 			<li class="🍺"><emo>🍺</emo> Ramble too much? Too verbose? Use <a href="https://huggingface.co/sleepdeprived3/Mistral-V7-Tekken-Concise">Mistral-V7-Tekken-Concise prompt</a></li>
-			<li class="🍺"><emo>🍺</emo> Guide to selecting the correct Mistral template</li>
+			<li class="🍺"><emo>🍺</emo> Guide to selecting the correct Mistral template 👇</li>
 
 		<li class="🤖" id="🤖Mistral Small 4"><b>Mistral Small 4</b></li>
 		  <li class="🍺">To enable reasoning, you need to connect via Chat Completion API</li>
@@ -353,8 +359,8 @@
 		<li class="▶️"><emo>💭</emo> Reasoning usage </li>
 			<li class="▶️▶️ ⚙️">Temperature 0.7</li>
 			<li class="▶️▶️ ⚙️">Reasoning_Effort High</li>
-			<li class="🍺"><emo>💭</emo> Reasoning Formatting: [THINK] [/THINK]</li>
-			<li class="🍺"><emo>🍺</emo> Guide to selecting the correct Mistral template</li>
+			<li class="▶️▶️ ⚙️"> Reasoning Formatting: [THINK] [/THINK]</li>
+			<li class="🍺"><emo>🍺</emo> Guide to selecting the correct Mistral template 👇</li>
 
 	<li class="🏢" id="🏢Nvidia"><i>Nvidia</i><flag>🇺🇸</flag></li>
 
@@ -384,7 +390,6 @@
 			<li class="⚙️">Temperature 0.6</li>
 			<li class="⚙️">Top_P 0.95</li>
 			<li class="⚙️">Only support 'Chat Completion API'</li>
-			<li class="⚙️"><emo>💭</emo> Reasoning formatting: &lt;think&gt;&lt;/think&gt;</li>
 			<li class="🔞"><emo>🔞</emo><emo>💥</emo> Disabling <emo>💭</emo>Reasoning prevents hard refusals, but decrease realism.</li>
 
 	<li class="🏢" id="🏢Microsoft"><i>Microsoft</i><flag>🇺🇸</flag></li>