chore(model gallery): 🤖 add 1 new models via gallery agent

gallery/index.yaml

@@ -3295,7 +3295,7 @@
     - gemma3
     - gemma-3
   overrides:
     #mmproj: gemma-3-27b-it-mmproj-f16.gguf
     parameters:
       model: gemma-3-27b-it-Q4_K_M.gguf
   files:
@@ -3313,7 +3313,7 @@
   description: |
     google/gemma-3-12b-it is an open-source, state-of-the-art, lightweight, multimodal model built from the same research and technology used to create the Gemini models. It is capable of handling text and image input and generating text output. It has a large context window of 128K tokens and supports over 140 languages. The 12B variant has been fine-tuned using the instruction-tuning approach. Gemma 3 models are suitable for a variety of text generation and image understanding tasks, including question answering, summarization, and reasoning. Their relatively small size makes them deployable in environments with limited resources such as laptops, desktops, or your own cloud infrastructure.
   overrides:
     #mmproj: gemma-3-12b-it-mmproj-f16.gguf
     parameters:
       model: gemma-3-12b-it-Q4_K_M.gguf
   files:
@@ -3331,7 +3331,7 @@
   description: |
     Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models. Gemma 3 models are multimodal, handling text and image input and generating text output, with open weights for both pre-trained variants and instruction-tuned variants. Gemma 3 has a large, 128K context window, multilingual support in over 140 languages, and is available in more sizes than previous versions. Gemma 3 models are well-suited for a variety of text generation and image understanding tasks, including question answering, summarization, and reasoning. Their relatively small size makes it possible to deploy them in environments with limited resources such as laptops, desktops or your own cloud infrastructure, democratizing access to state of the art AI models and helping foster innovation for everyone. Gemma-3-4b-it is a 4 billion parameter model.
   overrides:
     #mmproj: gemma-3-4b-it-mmproj-f16.gguf
     parameters:
       model: gemma-3-4b-it-Q4_K_M.gguf
   files:
@@ -6949,7 +6949,7 @@
       sha256: 2756551de7d8ff7093c2c5eec1cd00f1868bc128433af53f5a8d434091d4eb5a
       uri: huggingface://Triangle104/Nano_Imp_1B-Q8_0-GGUF/nano_imp_1b-q8_0.gguf
 - &qwen25
   name: "qwen2.5-14b-instruct" ## Qwen2.5
   icon: https://avatars.githubusercontent.com/u/141221163
   url: "github:mudler/LocalAI/gallery/chatml.yaml@master"
   license: apache-2.0
@@ -23203,3 +23203,44 @@
     - filename: Spiral-Qwen3-4B-Multi-Env.Q4_K_M.gguf
       sha256: e91914c18cb91f2a3ef96d8e62a18b595dd6c24fad901dea639e714bc7443b09
       uri: huggingface://mradermacher/Spiral-Qwen3-4B-Multi-Env-GGUF/Spiral-Qwen3-4B-Multi-Env.Q4_K_M.gguf
+- !!merge <<: *qwen3vl
+  name: "safework-rm-value-72b-i1"
+  urls:
+    - https://huggingface.co/mradermacher/SafeWork-RM-Value-72B-i1-GGUF
+  description: |
+    **Model Name:** SafeWork-RM-Value-72B
+    **Base Model:** Qwen2.5-VL-72B-Instruct
+    **Type:** Multimodal Reward Model (Value Verifier)
+    **License:** Apache 2.0
+    **Repository:** [AI45Research/SafeWork-RM-Value-72B](https://huggingface.co/AI45Research/SafeWork-RM-Value-72B)
+
+    ---
+
+    ### 📌 Description:
+
+    SafeWork-RM-Value-72B is a large-scale multimodal reward model developed as part of the **SafeWork-R1** framework, designed to evaluate and align AI responses with human values through intrinsic safety reasoning. Built upon the Qwen2.5-VL-72B-Instruct foundation, this model serves as a **value verifier** in the SafeLadder reinforcement learning framework, enabling the co-evolution of safety and general intelligence.
+
+    Unlike traditional RLHF methods that rely solely on human preferences, SafeWork-RM-Value-72B is trained with curated datasets focused on safety, moral reasoning, and factual verification, allowing it to develop deep self-reflection and reasoning capabilities. It assesses whether a given response (text + image) aligns with human values by analyzing reasoning steps internally and producing a final judgment in the form of `boxed{good}` or `boxed{bad}`.
+
+    ### 🎯 Key Features:
+    - **Multimodal:** Accepts both text and image inputs.
+    - **High Accuracy:** Outperforms several state-of-the-art models on value alignment benchmarks, especially when reasoning is enabled.
+    - **Self-Reflective Reasoning:** Encourages internal thinking before judgment, improving consistency and reliability.
+    - **Open & Transparent:** Released under Apache 2.0 license with full technical documentation and inference code.
+
+    ### 🧪 Use Case:
+    Ideal for fine-tuning or evaluating AI agents where safety, ethical reasoning, and value alignment are critical—such as in content moderation, responsible AI deployment, or reinforcement learning systems.
+
+    ### 📚 Learn More:
+    - [Technical Report (arXiv)](https://arxiv.org/abs/2507.18576)
+    - [GitHub Repository](https://github.com/AI45Lab/SafeWork-R1)
+    - [Online Demo](https://safework-r1.ai45.shlab.org.cn/)
+
+    > ✅ *Use this model to build safer, more trustworthy AI systems — grounded in both performance and ethics.*
+  overrides:
+    parameters:
+      model: SafeWork-RM-Value-72B.i1-Q4_K_M.gguf
+  files:
+    - filename: SafeWork-RM-Value-72B.i1-Q4_K_M.gguf
+      sha256: 8143fe81a7bd0e31a53e36ceb4cac69b9c1735b9e08dd1a7a726fc9735be92e1
+      uri: huggingface://mradermacher/SafeWork-RM-Value-72B-i1-GGUF/SafeWork-RM-Value-72B.i1-Q4_K_M.gguf