# Learning to Ideate for Scientific Impact

arXiv: https://arxiv.org/abs/2609.29802 · HTML: https://mlsanity.com/paper/2609.29802/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 20 of its 46 references that are in the map.
- First posted on arXiv 24 Sep 2026.

## Builds on (most cited first, 20 of 20)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researchers](https://mlsanity.com/paper/2409.04109/) (2024)
- [SciMON: Scientific Inspiration Machines Optimized for Novelty](https://mlsanity.com/paper/2305.14259/) (2023)
- [Towards End-to-End Automation of AI Research](https://mlsanity.com/paper/2606.15497/) (2026)
- [A Comprehensive Survey of Scientific Large Language Models and Their Applications in Scientific Discovery](https://mlsanity.com/paper/2406.10833/) (2024)
- [Deep Neural Networks for Rank-Consistent Ordinal Regression Based On Conditional Probabilities](https://mlsanity.com/paper/2111.08851/) (2021)
- [AI Idea Bench 2025: AI Research Idea Generation Benchmark](https://mlsanity.com/paper/2504.14191/) (2025)
- [Scientific production in the era of Large Language Models](https://mlsanity.com/paper/2601.13187/) (2026)
- [Forecasting high-impact research topics via machine learning on evolving knowledge graphs](https://mlsanity.com/paper/2402.08640/) (2024)
- [Equilibrate RLHF: Towards Balancing Helpfulness-Safety Trade-off in Large Language Models](https://mlsanity.com/paper/2502.11555/) (2025)
- [IRIS: Interactive Research Ideation System for Accelerating Scientific Discovery](https://mlsanity.com/paper/2504.16728/) (2025)
- [HindSight: Evaluating LLM-Generated Research Ideas via Future Impact](https://mlsanity.com/paper/2603.15164/) (2026)
- [A review of the literature on citation impact indicators](https://mlsanity.com/paper/1507.02099/) (2015)
- [From Newborn to Impact: Bias-Aware Citation Prediction](https://mlsanity.com/paper/2510.19246/) (2025)
- [Learning to Predict Future-Aligned Research Proposals with Language Models](https://mlsanity.com/paper/2603.27146/) (2026)
- [Proof of Time: A Benchmark for Evaluating Scientific Idea Judgments](https://mlsanity.com/paper/2601.07606/) (2026)
- [DeepInnovator: Triggering the Innovative Capabilities of LLMs](https://mlsanity.com/paper/2602.18920/) (2026)
- [LDC: Learning to Generate Research Idea with Dynamic Control](https://mlsanity.com/paper/2412.14626/) (2024)
- [SciImpact: A Multi-Dimensional, Multi-Field Benchmark for Scientific Impact Prediction](https://mlsanity.com/paper/2604.17141/) (2026)

## Abstract (arXiv)

Scientific ideation is increasingly mediated by large language models, but current ideation systems are usually trained and evaluated on immediately judgeable proxies such as novelty, clarity, and feasibility. This leaves open whether delayed signals of scientific uptake can be used as feedback for steering models toward research directions with higher expected impact. We study this question using citation-normalized impact as a noisy but scalable proxy for scholarly uptake. We construct a large-scale dataset from over 100K computer science papers by extracting goal-conditioned idea descriptions and assigning each paper an ordinal, year-normalized citation label. We then train a goal-conditioned reward model to predict citation-impact labels from research goal and idea pairs, and use this reward to align an idea generator through supervised fine-tuning followed by reinforcement learning. To reduce circularity, we evaluate generated ideas with a held-out, reference-grounded protocol that compares model outputs against historical ideas under the same research goal and weights judgments by the reference idea's citation-impact label. Experiments show that our RL-tuned model consistently produces ideas with higher estimated impact than both the base model and supervised fine-tuning baselines. Our findings position scientific impact as a practical, outcome-grounded feedback signal for aligning LLMs in open-ended scientific discovery.