@ModelScope2022: SciJudge-30B and 4B learn to predict which scientific work will carry stronger citation impact. License: Apache-2.0 30B…
Summary
ModelScope releases SciJudge-30B and SciJudge-4B, models trained on millions of arXiv papers to predict citation impact, achieving state-of-the-art accuracy surpassing larger models like GPT-5.2 and Gemini 3 Pro.
View Cached Full Text
Cached at: 07/03/26, 02:29 AM
SciJudge-30B and 4B learn to predict which scientific work will carry stronger citation impact. License: Apache-2.0
30B https://modelscope.ai/models/openmoss/SciJudge-30B-2605… 4B https://modelscope.ai/models/openmoss/SciJudge-4B-2605… https://modelscope.ai/papers/2603.14473…
Scientific Judge accuracy: SciJudge-30B reaches 80.6 in-domain, surpassing GPT-5.2, GLM-5, and Gemini 3 Pro; SciJudge-4B also outperforms much larger baseline models
Data signal: built from 2.1M arXiv papers and 696,758 field- and time-matched citation-based preference pairs
Training: GRPO with DAPO loss and citation-based pairwise rewards
Similar Articles
@AlphaSignalAI: A 4B model can now anticipate scientific breakthroughs before scientists do. Researchers often build breakthroughs by c…
A new paper introduces GIANTS-4B, a 4-billion-parameter model trained with reinforcement learning to predict scientific insights by combining ideas from foundational papers, achieving higher similarity and citation potential than larger models like Gemini 3 Pro.
@nickscamara_: New discoveries are gonna come from models that can reason over the latest science The rate of scientific progress beco…
Firecrawl released a state-of-the-art research index for AI/ML papers, claiming 18% better recall on arXivQA than competitors, designed for autonomous research agents.
PaperScorer: How Predictable Is Paper Acceptance?
The paper introduces PaperScorer, a family of models that predict a paper's peer-review accept/reject decision from text or page screenshots, achieving 68% accuracy on ICLR papers and 77% on arXiv ML papers by fine-tuning on anonymized corpora and outperforming frontier models like GPT and Gemini.
MOOSE-Star (ICML 2026): 7B model + 108K-paper dataset for scientific hypothesis discovery
MOOSE-Star presents a 7B model fine-tuned from DeepSeek-R1-Distill-Qwen-7B for scientific hypothesis discovery, along with a dataset of 108K NCBI papers. The model achieves state-of-the-art inspiration retrieval accuracy, outperforming larger models like GPT-5.4 and Gemini-3 Pro.
@MaximeRivest: wow, can't wait to try it. we can do so much in specialized 8b!
Ai2（@allen_ai）发布开源模型 AstaBrief 8B，可将复杂研究问题与文献片段转化为带引用的报告，支持本地运行，开放权重与训练数据；@MaximeRivest 表示期待尝试小型专用 8B 模型的潜力。