Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
We propose Reinforcement Learning from Community Feedback (RLCF), a training paradigm that uses large-scale community signals as supervision, and formulate scientific taste learning as a preference modeling and alignment problem.
| Date | Stars |
|---|---|
| 2026-07-31 | 427 |
| 2026-08-06 | 428 |
Today
+1 stars today
This week
— stars this week
This month
— stars this month
Momentum
4.0
growth rate 0.00%/day
<div align="center">
# AI Can Learn Scientific Taste
<a href='https://arxiv.org/abs/2603.14473'><img src='https://img.shields.io/badge/arXiv-2603.14473-b31b1b.svg'></a>
<a href='https://huggingface.co/collections/OpenMOSS-Team/ai-can-learn-scientific-taste'><img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Models%20%26%20Datasets-yellow'></a>
<a href='https://huggingface.co/papers/2603.14473'><img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Paper-yellow'></a>
<a href='https://tongjingqi.github.io/AI-Can-Learn-Scientific-Taste/'><img src='https://img.shields.io/badge/🌐%20Project-Page-blue'></a>
<a href='https://paperank.open-moss.com/'><img src='https://img.shields.io/badge/🎮%20Demo-Online-orange'></a>
<a href='LICENSE'><img src='https://img.shields.io/badge/License-Apache%202.0-green.svg'></a>
English | [中文](README_zh.md)
</div>
<div align="center">
<a href="https://huggingface.co/papers/2603.14473">
<img src="assets/huggingface_paper_gold_day.svg"/>
</a>
</div>
## 🎊 News
- [2026.07] 📊 The 2025 Future-Year benchmark and evaluation results have been refreshed (904 pairs).
- [2026.07] 🤗 The SciJudge-2506 series has been updated on Hugging Face. [[Collection](https://huggingface.co/collections/OpenMOSS-Team/ai-can-learn-scientific-taste)]
- [2026.03] 🎮 Online Demo is now available! [[Demo](https://paperank.open-moss.com/)]
- [2026.03] 📄 Paper available on arXiv. [[arXiv](https://arxiv.org/abs/2603.14473)]
- [2026.03] 🤗 Models released on Hugging Face. [[Collection](https://huggingface.co/collections/OpenMOSS-Team/ai-can-learn-scientific-taste)]
## 📌 Table of Contents
- [Abstract](#-abstract)
- [Overview of RLCF](#️-overview-of-rlcf)
- [Core Components](#-core-components)
- [Key Results](#-key-results)
- [Citation](#-citation)
- [License](#️-license)
## 📜 Abstract
Scientific discovery depends on expert judgement and foresight, which we call scientific taste: the ability to judge and propose research ideas with potential for long-term scientific impact.
Whether AI can learn this ability remains an open question. Here we provide evidence that artificial intelligence can learn judgement and ideation.
We introduce Reinforcement Learning from Community Feedback (RLCF), a training paradigm that uses large-scale signals from the scientific community as supervision. We first train Scientific Judge on field- and time-matched pairs of high- vs. low-citation papers to judge ideas. We then train a Scientific Thinker to propose research ideas with high potential impact.
Experiments show that the 30B Scientific Judge variant outperforms strong LLM baselines (e.g., GPT-5.4 Thinking), while Scientific Judge generalizes across future-year papers, unseen fields, and other community metrics. Furthermore, Scientific Thinker proposes research ideas with higher potential impact than baselines. These results suggest that AI can learn scientific taste, marking an important step towards AI systems that could help accelerate scientific discovery.
<div align="center">
<img src="assets/overview.png" width="100%" alt="Overview of Reinforcement Learning from Community Feedback" />
</div>
## 🏗️ Overview of RLCF
RLCF consists of three stages:
1. **Construct community preference**
Citations are converted into pairwise preference signals by matching papers within the same field and publication period.
2. **Preference modeling with Scientific Judge**
We train a generative reward model with GRPO that reasons over a pair of paper abstracts and predicts which one is more likely to have higher impact.
3. **Preference alignment with Scientific Thinker**
We use Scientific Judge as a reward model and optimize a policy model with comparison-based GRPO to generate higher-impact research ideas.
## 🧩 Core Components
### Scientific Judge
- A generative reward model that reasons over paired paper abstracts and predicts which has higher potential impact.
- Trained with GRPO on 720K field- Excerpt of 6,629 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:64744bed76297fda, topic:rl, desc:reinforcement learning