Scaling Scientific Discovery Environments for Turn-Level Agentic RL

• URL: http://arxiv.org/abs/2607.28990v1
• Date: Fri, 31 Jul 2026 03:34:39 GMT
• Title: Scaling Scientific Discovery Environments for Turn-Level Agentic RL
• Abstract summary: This paper introduces SciDisco, a framework for training Scientific Discovery agents in process-verifiable environments.<n>SciThque compiles hypotheses, datasets, hidden evidence graphs, and verifiers into task environments where analytical progress can be checked.<n>Experiments show that SciDisco-14B reaches state-of-the-art on hypothesis-driven scientific data analysis benchmarks.
• Score: 14.746461766950162
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language model agents have shown promising capabilities in data-driven scientific discovery tasks, where an agent interacts with an execution environment and produces a statistical claim. Long-horizon scientific analysis remains constrained by the lack of process supervised environments over real-world scientific data. This paper introduces SciDisco, a scalable framework for training Scientific Discovery agents in process-verifiable environments. SciThèque compiles hypotheses, datasets, hidden evidence graphs, and verifiers into task environments where analytical progress can be checked during interaction. DAG-grounded trajectory synthesis uses these environments to construct verifier-filtered multi-turn demonstrations. DiscoPO then uses the environment as the source of training signal, assigning turn-level credit to actions that produce verifiable analytical evidence. Experiments show that SciDisco-14B reaches state-of-the-art on hypothesis-driven scientific data analysis benchmarks.

Related papers

• EarthVerse: Benchmarking Scientific Agents Across Dynamic Earth Systems and Natural Hazards [73.12811119273206]
  We introduce EarthVerse, a benchmark that evaluates scientific agents through package-scoped investigations.<n>Agents inspect heterogeneous event packages, choose compatible evidence, execute transparent calculations, reconcile source differences, and preserve provenance in the final answer.<n>Across systems, the best mean answer-unit accuracy is 84.65%, while the highest Strict@95 is only 34.81%.
  arXiv  Detail & Related papers  (2026-08-24T17:29:16Z)
• Rethinking Scientific Discovery in the Agentic Era [69.85050591628048]
  SCION is an agentic scientific operating system that acts as an textbf organizational nexus.<n>At its core is the textbfResearch Execution Plan (REP), which compiles high-level scientific intent into staged objectives, dependencies, verification checkpoints, tool requirements, expected artifacts, and fallback conditions.<n>We formulate discovery under SCION as textbfTarget-conditioned Inverse Search and extend it to hidden-target settings through batch active search under finite experimental budgets.
  arXiv  Detail & Related papers  (2026-07-04T13:09:33Z)
• DiffUNet^2: Bidirectional Prediction, Probabilistic Generation and Collaborative Visual Discovery for Scientific Data [25.698658042998073]
  temporal evolution is important to analyzing and reasoning about scientific phenomena.<n>Most machine learning methods provide deterministic forward predictions that overlook multiple plausible outcomes.<n>We present a framework that integrates diffusion-based generative modeling with interactive visual analytics for scientific exploration.
  arXiv  Detail & Related papers  (2026-06-02T17:15:01Z)
• Toward a Scientific Discovery Engine for Weather and Climate Data: A Visual Analytics Workbench for Embedding-Based Exploration [0.0]
  Earth system science is producing increasingly large, high-dimensional datasets from physics based Earth system models to AI-based weather and climate models.<n> Embedding-based representations can make these data searchable through similarity search and analog retrieval, but nearest neighbors in latent space are not automatically scientifically meaningful.<n>We present an open-source visual analytics workbench for each of these steps.
  arXiv  Detail & Related papers  (2026-05-01T17:03:33Z)
• Causal Representation Meets Stochastic Modeling under Generic Geometry [49.24293444627916]
  We develop causal representation learning for continuous-time latent point processes.<n>We develop MUTATE, an identifiable variational autoencoder framework with a time-adaptive transition module.<n>Across simulated and empirical studies, we find that MUTATE can effectively answer scientific questions.
  arXiv  Detail & Related papers  (2026-02-04T20:40:53Z)
• WildSci: Advancing Scientific Reasoning from In-the-Wild Literature [50.16160754134139]
  We introduce WildSci, a new dataset of domain-specific science questions automatically synthesized from peer-reviewed literature.<n>By framing complex scientific reasoning tasks in a multiple-choice format, we enable scalable training with well-defined reward signals.<n>Experiments on a suite of scientific benchmarks demonstrate the effectiveness of our dataset and approach.
  arXiv  Detail & Related papers  (2026-01-09T06:35:23Z)
• SpatialBench: Can Agents Analyze Real-World Spatial Biology Data? [6.993633248897315]
  We introduce SpatialBench, a benchmark of 146 verifiable problems derived from practical spatial analysis.<n>Each problem provides a snapshot of experimental data immediately prior to an analysis step.<n>Base model accuracy remains low, with strong model-task and model-platform interactions.
  arXiv  Detail & Related papers  (2025-12-26T07:40:11Z)
• A Survey of Scientific Large Language Models: From Data Foundations to Agent Frontiers [251.23085679210206]
  Scientific Large Language Models (Sci-LLMs) are transforming how knowledge is represented, integrated, and applied in scientific research.<n>This survey reframes the development of Sci-LLMs as a co-evolution between models and their underlying data substrate.<n>We formulate a unified taxonomy of scientific data and a hierarchical model of scientific knowledge.
  arXiv  Detail & Related papers  (2025-08-28T18:30:52Z)
• Operationalizing Serendipity: Multi-Agent AI Workflows for Enhanced Materials Characterization with Theory-in-the-Loop [0.0]
  SciLink is an open-source, multi-agent artificial intelligence framework designed to operationalize serendipity in materials research.<n>It creates a direct, automated link between experimental observation, novelty assessment, and theoretical simulations.<n>We show its application to atomic-resolution and hyperspectral data, its capacity to integrate real-time human expert guidance, and its ability to close the research loop.
  arXiv  Detail & Related papers  (2025-08-07T04:59:17Z)
• Causal Representation Learning in Temporal Data via Single-Parent Decoding [66.34294989334728]
  Scientific research often seeks to understand the causal structure underlying high-level variables in a system. Scientists typically collect low-level measurements, such as geographically distributed temperature readings. We propose a differentiable method, Causal Discovery with Single-parent Decoding, that simultaneously learns the underlying latents and a causal graph over them.
  arXiv  Detail & Related papers  (2024-10-09T15:57:50Z)

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.