Update README.md

README.md

@@ -24,4 +24,24 @@ tags:
 
 # **MedChain-R1-9B-GGUF**
 
-> **MedChain-R1-9B** is a 9B vision-language model for sequential medical image grounding, built on Qwen3.5-9B and described in "Think Before You Locate: Clinical Chain-of-Thought Supervision for Sequential Medical Image Grounding" (arXiv:2611.00041). Given a sequence of medical images and a clinical question, it first writes an image-grounded rationale in `<think>` tags, then returns one pixel-coordinate bounding box. Training used deliberative chain-of-thought supervision (dCoT-SFT) on the MedChain corpus, followed by GRPO with an IoU plus format reward. MedChain has 182,573 filtered rationales over the 8 MedSG-Bench tasks and 10 imaging modalities. A medical VLM (FlemingVL-38B) drafts each rationale and a multimodal refiner (Qwen3.5-397B-A17B) rewrites it conditioned on the annotated box. This checkpoint scores 77.54 mIoU / 84.87 Acc@0.5 on all 9,630 MedSG-Bench test questions, above the paper's 9B GRPO row (76.90 / 83.82), which came from a different checkpoint. It also beats the supervised state-of-the-art MedSeq-Grounder 7B (72.55 / 79.71) and is far ahead of the zero-shot general and medical baselines, which mostly score below 21 mIoU. The paper notes that the gains vary by task. Against an answer-only 4B model, the reasoning supervision gains 14.7–16.1 mIoU on difference grounding but loses 2.2–3.2 on multi-view, patch and referring grounding. It adds that GRPO contributes little after dCoT-SFT (+0.37 mIoU at 4B). The card says to generate one example at a time, because batched generation gave degenerate outputs, and to use the bundled processor. The rationales were written knowing the answer, so they may be post-hoc rather than a faithful record of the model's reasoning. The model is a research artifact, not validated for diagnosis or clinical deployment, and evaluation is limited to MedSG-Bench. It is released under Apache 2.0.
+> **MedChain-R1-9B** is a 9B vision-language model for sequential medical image grounding, built on Qwen3.5-9B and described in "Think Before You Locate: Clinical Chain-of-Thought Supervision for Sequential Medical Image Grounding" (arXiv:2611.00041). Given a sequence of medical images and a clinical question, it first writes an image-grounded rationale in `<think>` tags, then returns one pixel-coordinate bounding box. Training used deliberative chain-of-thought supervision (dCoT-SFT) on the MedChain corpus, followed by GRPO with an IoU plus format reward. MedChain has 182,573 filtered rationales over the 8 MedSG-Bench tasks and 10 imaging modalities. A medical VLM (FlemingVL-38B) drafts each rationale and a multimodal refiner (Qwen3.5-397B-A17B) rewrites it conditioned on the annotated box. This checkpoint scores 77.54 mIoU / 84.87 Acc@0.5 on all 9,630 MedSG-Bench test questions, above the paper's 9B GRPO row (76.90 / 83.82), which came from a different checkpoint. It also beats the supervised state-of-the-art MedSeq-Grounder 7B (72.55 / 79.71) and is far ahead of the zero-shot general and medical baselines, which mostly score below 21 mIoU. The paper notes that the gains vary by task. Against an answer-only 4B model, the reasoning supervision gains 14.7–16.1 mIoU on difference grounding but loses 2.2–3.2 on multi-view, patch and referring grounding. It adds that GRPO contributes little after dCoT-SFT (+0.37 mIoU at 4B). The card says to generate one example at a time, because batched generation gave degenerate outputs, and to use the bundled processor. The rationales were written knowing the answer, so they may be post-hoc rather than a faithful record of the model's reasoning. The model is a research artifact, not validated for diagnosis or clinical deployment, and evaluation is limited to MedSG-Bench. It is released under Apache 2.0.
+
+## Model Files
+
+| File Name | Quant Type | File Size | File Link | Description |
+|-----------|------------|-----------|-----------|-------------|
+| MedChain-R1-9B.BF16.gguf | BF16 | 17.9 GB | [Link](https://huggingface.co/prithivMLmods/MedChain-R1-9B-GGUF/blob/main/MedChain-R1-9B.BF16.gguf) | Full BF16 weights. Highest quality, largest file size. |
+| MedChain-R1-9B.Q3_K_L.gguf | Q3_K_L | 4.93 GB | [Link](https://huggingface.co/prithivMLmods/MedChain-R1-9B-GGUF/blob/main/MedChain-R1-9B.Q3_K_L.gguf) | Lower quality but usable, good for low RAM availability. |
+| MedChain-R1-9B.Q3_K_M.gguf | Q3_K_M | 4.62 GB | [Link](https://huggingface.co/prithivMLmods/MedChain-R1-9B-GGUF/blob/main/MedChain-R1-9B.Q3_K_M.gguf) | Low quality. |
+| MedChain-R1-9B.Q4_K_M.gguf | Q4_K_M | 5.63 GB | [Link](https://huggingface.co/prithivMLmods/MedChain-R1-9B-GGUF/blob/main/MedChain-R1-9B.Q4_K_M.gguf) | Good quality, default size for most use cases, *recommended*. |
+| MedChain-R1-9B.Q4_K_S.gguf | Q4_K_S | 5.35 GB | [Link](https://huggingface.co/prithivMLmods/MedChain-R1-9B-GGUF/blob/main/MedChain-R1-9B.Q4_K_S.gguf) | Slightly lower quality with more space savings, *recommended*. |
+| MedChain-R1-9B.Q5_K_M.gguf | Q5_K_M | 6.47 GB | [Link](https://huggingface.co/prithivMLmods/MedChain-R1-9B-GGUF/blob/main/MedChain-R1-9B.Q5_K_M.gguf) | High quality, *recommended*. |
+| MedChain-R1-9B.Q5_K_S.gguf | Q5_K_S | 6.31 GB | [Link](https://huggingface.co/prithivMLmods/MedChain-R1-9B-GGUF/blob/main/MedChain-R1-9B.Q5_K_S.gguf) | High quality, *recommended*. |
+| MedChain-R1-9B.Q6_K.gguf | Q6_K | 7.36 GB | [Link](https://huggingface.co/prithivMLmods/MedChain-R1-9B-GGUF/blob/main/MedChain-R1-9B.Q6_K.gguf) | Very high quality, near perfect, *recommended*. |
+| MedChain-R1-9B.mmproj-bf16.gguf | mmproj-bf16 | 922 MB | [Link](https://huggingface.co/prithivMLmods/MedChain-R1-9B-GGUF/blob/main/MedChain-R1-9B.mmproj-bf16.gguf) | Multimodal projection file in BF16 format. Used for vision/language models. |
+
+## llama.cpp
+
+LLM inference in C/C++ — [https://github.com/ggml-org/llama.cpp](https://github.com/ggml-org/llama.cpp)
+
+Quantized using llama.cpp by [prithivMLmods](https://huggingface.co/prithivMLmods).