AI Scientist via Synthetic Task Scaling

• URL: http://arxiv.org/abs/2603.17216v1
• Date: Tue, 17 Mar 2026 23:43:16 GMT
• Title: AI Scientist via Synthetic Task Scaling
• Abstract summary: We provide a novel synthetic environment generation pipeline targeting machine learning agents.<n>Our pipeline automatically synthesizes machine learning challenges compatible with the SWE-agent framework.<n>We tackle MLGym, a benchmark for machine learning tasks.
• Score: 2.056976961583532
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: With the advent of AI agents, automatic scientific discovery has become a tenable goal. Many recent works scaffold agentic systems that can perform machine learning research, but don't offer a principled way to train such agents -- and current LLMs often generate plausible-looking but ineffective ideas. To make progress on training agents that can learn from doing, we provide a novel synthetic environment generation pipeline targeting machine learning agents. Our pipeline automatically synthesizes machine learning challenges compatible with the SWE-agent framework, covering topic sampling, dataset proposal, and code generation. The resulting synthetic tasks are 1) grounded in real machine learning datasets, because the proposed datasets are verified against the Huggingface API and are 2) verified for higher quality with a self-debugging loop. To validate the effectiveness of our synthetic tasks, we tackle MLGym, a benchmark for machine learning tasks. From the synthetic tasks, we sample trajectories from a teacher model (GPT-5), then use the trajectories to train a student model (Qwen3-4B and Qwen3-8B). The student models trained with our synthetic tasks achieve improved performance on MLGym, raising the AUP metric by 9% for Qwen3-4B and 12% for Qwen3-8B.

Related papers

• FrogNano: Training a 4B Coding Agent via Online Task Synthesis [61.25325155871018]
  FrogNano is a 4B coding agent designed to tackle software engineering (SWE) tasks efficiently and effectively.<n>It is post-trained exclusively via RL on around 1,500 SWE environments with synthetic tasks.
  arXiv  Detail & Related papers  (2026-09-07T19:39:38Z)
• Environment-free Synthetic Data Generation for API-Calling Agents [68.95520739708304]
  Training API-calling large language model (LLM) agents demands massive amounts of high-quality trajectories.<n>We propose an environment-free synthetic data generation approach that leverages LLMs as on-the-fly digital world models.<n>We evaluate our approach on the challenging AppWorld and OfficeBench benchmarks, which include both information-retrieval and state-changing tasks.
  arXiv  Detail & Related papers  (2026-07-18T17:36:44Z)
• AgenticQwen: Training Small Agentic Language Models with Dual Data Flywheels for Industrial-Scale Tool Use [13.583197273673974]
  We introduce the AgenticQwen family of models, trained via multi-round reinforcement learning (RL) on synthetic data and a limited amount of open-source data.<n>Our training framework combines reasoning RL and agentic RL with dual data flywheels that automatically generate increasingly challenging tasks.<n>The models achieve strong performance on multiple agentic benchmarks, and in our industrial agent system, close the gap with much larger models on search and data analysis tasks.
  arXiv  Detail & Related papers  (2026-04-23T12:14:52Z)
• Optimizing Neurorobot Policy under Limited Demonstration Data through Preference Regret [45.21845762562799]
  "Master your own expertise" (MYOE) is a self-imitation framework that enables robotic agents to learn complex behaviors from limited demonstration data samples.<n>Inspired by human perception and action, we propose and design what we call the queryable mixture-of-preferences state space model (QMoP-SSM)<n>Our experiments demonstrate the robustness, adaptability, and out-of-sample performance of our agent compared to other state-of-the-art RLfD schemes.
  arXiv  Detail & Related papers  (2026-04-04T00:03:59Z)
• Learning Adaptive Force Control for Contact-Rich Sample Scraping with Heterogeneous Materials [0.5161531917413708]
  Increasing demand for accelerated scientific discovery, driven by global challenges, highlights the need for advanced AI-driven robotics.<n> deploying robotic chemists in human-centric labs is key for the next horizon of autonomous discovery.<n>Our work proposes an adaptive control framework to address this, relying on a low-level Cartesian impedance controller for stable and compliant physical interaction.<n>Our method was evaluated across five different material setups, outperforming a fixed-wrench baseline by an average of 10.9%.
  arXiv  Detail & Related papers  (2026-03-11T17:07:55Z)
• AIRS-Bench: a Suite of Tasks for Frontier AI Research Science Agents [49.67355440164857]
  We introduce AIRS-Bench, a suite of 20 tasks sourced from state-of-the-art machine learning papers.<n>Airs-Bench tasks assess agentic capabilities over the full research lifecycle.<n>We open-source the AIRS-Bench task definitions and evaluation code to catalyze further development in autonomous scientific research.
  arXiv  Detail & Related papers  (2026-02-06T16:45:02Z)
• AnyTask: an Automated Task and Data Generation Framework for Advancing Sim-to-Real Policy Learning [16.837846476054786]
  Generalist robot learning remains constrained by data: large-scale, diverse, and high-quality interaction data are expensive to collect in the real world.<n>We present AnyTask, an automated framework that pairs massively parallel GPU simulation with foundation models to design diverse manipulation tasks.<n>We train behavior cloning policies on generated data, validate them in simulation, and deploy them directly on real robot hardware.
  arXiv  Detail & Related papers  (2025-12-19T17:55:48Z)
• Meta-Learning and Synthetic Data for Automated Pretraining and Finetuning [2.657867981416885]
  Growing number of pretrained models in Machine Learning (ML) presents significant challenges for practitioners.<n>As models grow in scale, the increasing reliance on real-world data poses a bottleneck for training and requires leveraging data more effectively.<n>This dissertation adopts meta-learning to extend automated machine learning to the deep learning domain.
  arXiv  Detail & Related papers  (2025-06-11T12:48:45Z)
• OS-Genesis: Automating GUI Agent Trajectory Construction via Reverse Task Synthesis [55.390060529534644]
  We propose OS-Genesis, a novel data synthesis pipeline for Graphical User Interface (GUI) agents.<n>Instead of relying on pre-defined tasks, OS-Genesis enables agents first to perceive environments and perform step-wise interactions.<n>We demonstrate that training GUI agents with OS-Genesis significantly improves their performance on highly challenging online benchmarks.
  arXiv  Detail & Related papers  (2024-12-27T16:21:58Z)
• Robot Fine-Tuning Made Easy: Pre-Training Rewards and Policies for Autonomous Real-World Reinforcement Learning [58.3994826169858]
  We introduce RoboFuME, a reset-free fine-tuning system for robotic reinforcement learning. Our insights are to utilize offline reinforcement learning techniques to ensure efficient online fine-tuning of a pre-trained policy. Our method can incorporate data from an existing robot dataset and improve on a target task within as little as 3 hours of autonomous real-world experience.
  arXiv  Detail & Related papers  (2023-10-23T17:50:08Z)
• MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation [96.71370747681078]
  We introduce MLAgentBench, a suite of 13 tasks ranging from improving model performance on CIFAR-10 to recent research problems like BabyLM. For each task, an agent can perform actions like reading/writing files, executing code, and inspecting outputs. We benchmark agents based on Claude v1.0, Claude v2.1, Claude v3 Opus, GPT-4, GPT-4-turbo, Gemini-Pro, and Mixtral and find that a Claude v3 Opus agent is the best in terms of success rate.
  arXiv  Detail & Related papers  (2023-10-05T04:06:12Z)
• Learning from synthetic data generated with GRADE [0.6982738885923204]
  We present a framework for generating realistic animated dynamic environments (GRADE) for robotics research. GRADE supports full simulation control, ROS integration, realistic physics, while being in an engine that produces high visual fidelity images and ground truth data. We show that, even training using only synthetic data, can generalize well to real-world images in the same application domain.
  arXiv  Detail & Related papers  (2023-05-07T14:13:04Z)
• Hindsight States: Blending Sim and Real Task Elements for Efficient Reinforcement Learning [61.3506230781327]
  In robotics, one approach to generate training data builds on simulations based on dynamics models derived from first principles. Here, we leverage the imbalance in complexity of the dynamics to learn more sample-efficiently. We validate our method on several challenging simulated tasks and demonstrate that it improves learning both alone and when combined with an existing hindsight algorithm.
  arXiv  Detail & Related papers  (2023-03-03T21:55:04Z)
• Learning Task Automata for Reinforcement Learning using Hidden Markov Models [37.69303106863453]
  This paper proposes a novel pipeline for learning non-Markovian task specifications as succinct finite-state task automata' We learn a product MDP, a model composed of the specification's automaton and the environment's MDP, by treating the product MDP as a partially observable MDP and using the well-known Baum-Welch algorithm for learning hidden Markov models. Our learnt task automaton enables the decomposition of a task into its constituent sub-tasks, which improves the rate at which an RL agent can later synthesise an optimal policy.
  arXiv  Detail & Related papers  (2022-08-25T02:58:23Z)

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.