Training AI Co-Scientists Using Rubric Rewards

• URL: http://arxiv.org/abs/2512.23707v1
• Date: Mon, 29 Dec 2025 18:59:33 GMT
• Title: Training AI Co-Scientists Using Rubric Rewards
• Authors: Shashwat Goel, Rishi Hazra, Dulhan Jayalath, Timon Willi, Parag Jain, William F. Shen, Ilias Leontiadis, Francesco Barbieri, Yoram Bachrach, Jonas Geiping, Chenxi Whitehouse,
• Abstract summary: A crucial feature of AI co-scientists is the ability to generate a research plan given a set of aims and constraints.<n>In this work, we study how to leverage the vast corpus of existing research papers to train language models that generate better research plans.<n>We build a scalable, diverse training corpus by automatically extracting research goals and goal-specific grading rubrics from papers across several domains.
• Score: 36.857905469178796
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: AI co-scientists are emerging as a tool to assist human researchers in achieving their research goals. A crucial feature of these AI co-scientists is the ability to generate a research plan given a set of aims and constraints. The plan may be used by researchers for brainstorming, or may even be implemented after further refinement. However, language models currently struggle to generate research plans that follow all constraints and implicit requirements. In this work, we study how to leverage the vast corpus of existing research papers to train language models that generate better research plans. We build a scalable, diverse training corpus by automatically extracting research goals and goal-specific grading rubrics from papers across several domains. We then train models for research plan generation via reinforcement learning with self-grading. A frozen copy of the initial policy acts as the grader during training, with the rubrics creating a generator-verifier gap that enables improvements without external human supervision. To validate this approach, we conduct a study with human experts for machine learning research goals, spanning 225 hours. The experts prefer plans generated by our finetuned Qwen3-30B-A3B model over the initial model for 70% of research goals, and approve 84% of the automatically extracted goal-specific grading rubrics. To assess generality, we also extend our approach to research goals from medical papers, and new arXiv preprints, evaluating with a jury of frontier models. Our finetuning yields 12-22% relative improvements and significant cross-domain generalization, proving effective even in problem settings like medical research where execution feedback is infeasible. Together, these findings demonstrate the potential of a scalable, automated training recipe as a step towards improving general AI co-scientists.

Related papers

• Step-DeepResearch Technical Report [90.50586290399683]
  We introduce Step-DeepResearch, a cost-effective, end-to-end agent.<n>We propose a Data Synthesis Strategy Based on Atomic Capabilities to reinforce planning and report writing.<n>To bridge the evaluation gap in the Chinese domain, we establish ADR-Bench for realistic deep research scenarios.
  arXiv  Detail & Related papers  (2025-12-23T16:32:27Z)
• From AutoRecSys to AutoRecLab: A Call to Build, Evaluate, and Govern Autonomous Recommender-Systems Research Labs [5.34658805289521]
  We argue for a shift from narrow AutoRecSys tools to an Autonomous Recommender-Systems Research Lab (AutoRecLab)<n>AutoRecLab integrates end-to-end automation: problem ideation, literature analysis, experimental design and execution, result interpretation, manuscript drafting, and logging.<n>We conclude with a call to organise a community retreat to coordinate next steps and co-author guidance for the responsible integration of automated research systems.
  arXiv  Detail & Related papers  (2025-10-20T20:58:50Z)
• AI, Humans, and Data Science: Optimizing Roles Across Workflows and the Workforce [0.0]
  We consider the potential and limitation of analytic, generative, and agentic AI to augment data scientists or take on tasks traditionally done by human analysts and researchers.<n>Just as earlier eras of survey analysis created issues when the increased ease of using statistical software allowed researchers to conduct analyses they did not fully understand, the new AI tools may create similar but larger risks.
  arXiv  Detail & Related papers  (2025-07-15T17:59:06Z)
• The Budget AI Researcher and the Power of RAG Chains [4.797627592793464]
  Current approaches to supporting research idea generation often rely on generic large language models (LLMs)<n>Our framework, The Budget AI Researcher, uses retrieval-augmented generation chains, vector databases, and topic-guided pairing to recombine concepts from hundreds of machine learning papers.<n>The system ingests papers from nine major AI conferences, which collectively span the vast subfields of machine learning, and organizes them into a hierarchical topic tree.
  arXiv  Detail & Related papers  (2025-06-14T02:40:35Z)
• From Automation to Autonomy: A Survey on Large Language Models in Scientific Discovery [67.07598263346591]
  Large Language Models (LLMs) are catalyzing a paradigm shift in scientific discovery.<n>This survey systematically charts this burgeoning field, placing a central focus on the changing roles and escalating capabilities of LLMs in science.
  arXiv  Detail & Related papers  (2025-05-19T15:41:32Z)
• MLGym: A New Framework and Benchmark for Advancing AI Research Agents [51.9387884953294]
  We introduce Meta MLGym and MLGym-Bench, a new framework and benchmark for evaluating and developing large language models on AI research tasks.<n>This is the first Gym environment for machine learning (ML) tasks, enabling research on reinforcement learning (RL) algorithms for training such agents.<n>We evaluate a number of frontier large language models (LLMs) on our benchmarks such as Claude-3.5-Sonnet, Llama-3.1 405B, GPT-4o, o1-preview, and Gemini-1.5 Pro.
  arXiv  Detail & Related papers  (2025-02-20T12:28:23Z)
• AIGS: Generating Science from AI-Powered Automated Falsification [17.50867181053229]
  We propose Baby-AIGS as a baby-step demonstration of a full-process AIGS system, which is a multi-agent system with agents in roles representing key research process. Experiments on three tasks preliminarily show that Baby-AIGS could produce meaningful scientific discoveries, though not on par with experienced human researchers.
  arXiv  Detail & Related papers  (2024-11-17T13:40:35Z)
• CycleResearcher: Improving Automated Research via Automated Review [37.03497673861402]
  This paper explores the possibility of using open-source post-trained large language models (LLMs) as autonomous agents capable of performing the full cycle of automated research and review.<n>To train these models, we develop two new datasets, reflecting real-world machine learning research and peer review dynamics.<n>Our results demonstrate that CycleReviewer achieves promising performance with a 26.89% reduction in mean absolute error (MAE) compared to individual human reviewers in predicting paper scores.
  arXiv  Detail & Related papers  (2024-10-28T08:10:21Z)
• Chain of Ideas: Revolutionizing Research Via Novel Idea Development with LLM Agents [64.64280477958283]
  An exponential increase in scientific literature makes it challenging for researchers to stay current with recent advances and identify meaningful research directions. Recent developments in large language models(LLMs) suggest a promising avenue for automating the generation of novel research ideas. We propose a Chain-of-Ideas(CoI) agent, an LLM-based agent that organizes relevant literature in a chain structure to effectively mirror the progressive development in a research domain.
  arXiv  Detail & Related papers  (2024-10-17T03:26:37Z)
• ResearchAgent: Iterative Research Idea Generation over Scientific Literature with Large Language Models [56.08917291606421]
  ResearchAgent is an AI-based system for ideation and operationalization of novel work.<n>ResearchAgent automatically defines novel problems, proposes methods and designs experiments, while iteratively refining them.<n>We experimentally validate our ResearchAgent on scientific publications across multiple disciplines.
  arXiv  Detail & Related papers  (2024-04-11T13:36:29Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.