Update SciJudge-30B model card and link 2605 release

README.md

@@ -1,32 +1,40 @@
 ---
 language:
-  - en
+- en
 license: apache-2.0
 base_model: Qwen/Qwen3-30B-A3B-Instruct-2507
+datasets:
+- OpenMOSS-Team/SciJudgeBench
 tags:
-  - scientific-evaluation
-  - citation-prediction
-  - preference-learning
-  - GRPO
-  - moe
+- scientific-taste
+- GRPO
 pipeline_tag: text-generation
 library_name: transformers
 ---
 
-# SciJudge-Qwen3-30B
+# SciJudge-30B
 
-SciJudge-Qwen3-30B is a fine-tuned language model for **scientific paper evaluation**. Given two academic papers' metadata (title, abstract, publication date), it predicts which paper has a higher citation count — serving as a proxy for assessing research impact and "scientific taste."
+> **Update:** A newer release is available at [SciJudge-30B-2605](https://huggingface.co/OpenMOSS-Team/SciJudge-30B-2605). We recommend using the newer release for current experiments and comparisons.
 
-This model is part of the paper: **[AI Can Learn Scientific Taste](https://arxiv.org/abs/2603.14473)**.
+SciJudge-30B is a Qwen3-30B-A3B-Instruct-2507 MoE model fine-tuned for scientific paper evaluation. Given two papers' titles, abstracts, and publication dates, it predicts which paper has higher citation impact.
+
+This model is part of [AI Can Learn Scientific Taste](https://arxiv.org/abs/2603.14473). The benchmark dataset is [SciJudgeBench](https://huggingface.co/datasets/OpenMOSS-Team/SciJudgeBench).
+
+Resources: [Project page](https://tongjingqi.github.io/AI-Can-Learn-Scientific-Taste/) and [GitHub repository](https://github.com/tongjingqi/AI-Can-Learn-Scientific-Taste).
 
 ## Usage
 
 ```python
+import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
 model_name = "OpenMOSS-Team/SciJudge-30B"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="bfloat16", device_map="auto")
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    torch_dtype=torch.bfloat16,
+    device_map="auto",
+)
 
 messages = [
     {"role": "system", "content": "You are a helpful assistant. You first think about the reasoning process in your mind and then provide the user with the answer."},
@@ -42,20 +50,22 @@ print(response)
 
 ## Training Details
 
-- **Base model:** Qwen3-30B-A3B-Instruct-2507 (MoE, 30B total / 3B active)
-- **Training method:** GRPO (Generative Reward Policy Optimization) with DAPO loss
-- **Training data:** 720,341 preference pairs from arXiv papers
-- **Learning rate:** 8e-7 (cosine schedule, 5% warmup)
-- **Micro batch size:** 8, global batch size: 1024
-- **Optimizer:** Adam (with CPU offload)
+- **Base model:** [Qwen/Qwen3-30B-A3B-Instruct-2507](https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507)
+- **Training method:** GRPO with DAPO loss
+- **Reward:** external preference reward for citation-based pairwise judgment
 - **Precision:** bfloat16
-- **KL coefficient (β):** 0.03
+- **KL coefficient:** 0.03
 
 ## Citation
 
 ```bibtex
-@article{scijudge2025,
-  title={AI Can Learn Scientific Taste},
-  year={2025}
+@misc{tong2026ailearnscientifictaste,
+      title={AI Can Learn Scientific Taste}, 
+      author={Jingqi Tong and Mingzhe Li and Hangcheng Li and Yongzhuo Yang and Yurong Mou and Weijie Ma and Zhiheng Xi and Hongji Chen and Xiaoran Liu and Qinyuan Cheng and Ming Zhang and Qiguang Chen and Weifeng Ge and Qipeng Guo and Tianlei Ying and Tianxiang Sun and Yining Zheng and Xinchi Chen and Jun Zhao and Ning Ding and Xuanjing Huang and Yugang Jiang and Xipeng Qiu},
+      year={2026},
+      eprint={2603.14473},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2603.14473}, 
 }
 ```