corral-reasoning-annotations

# *Corral* – Reasoning Annotations <div align="center">  [](https://lamalab-org.github.io/corral/) [](https://lamalab-org.github.io/corral/docs/) [](https://github.com/lamalab-org/corral) [](https://opensource.org/licenses/MIT) [](https://arxiv.org/abs/2604.18805) [](https://huggingface.co/datasets/jablonkagroup/corral-reasoning-annotations) LLM epistemic annotations over Corral traces where the annotator judged that the agents do not reason scientifically </div> --- ## 📋 Dataset Summary This dataset is part of the *Corral* collection accompanying the paper [*AI scientists produce results without reasoning scientifically*](https://arxiv.org/abs/2604.18805). It contains **annotated evaluation traces** with **LLM-generated epistemic annotations** across the *Corral* benchmark. The dataset is exposed as **three model-specific configurations**: `claude_sonnet_45`, `gpt_4o`, and `gpt_oss_120b`. Within each config, each row corresponds to one annotated trace instance. The included annotations were produced by an **LLM annotator** that identified these cases as ones where the agents **do not reason scientifically**. This resource is intended for auditing, qualitative analysis, and process-level study of scientific-agent behaviour rather than for general-purpose model pre-training. ### 🎯 Supported Uses - 🧠 Studying LLM-generated epistemic annotations over scientific-agent traces - 📊 Auditing cases where an automatic annotator flagged non-scientific reasoning - 📐 Comparing annotated reasoning failures across models and trace files - 🔁 Building qualitative analysis sets for reasoning-process studies in scientific agents --- ## 🧪 About *Corral* [*Corral*](https://lamalab-org.github.io/corral/) is a framework for the *science of agents and agents for science*. It provides a microservice architecture that **decouples agents from environments** via a client–server design (REST API), ensuring flexibility, reproducibility, and robust isolation. - 🌍 **Environments** define the task space, available tools, and observable feedback — from chemistry labs to HPC clusters. - 🤖 **Agents** are modular LLM-based entities supporting scaffolds such as ReAct, ToolCalling, LLMPlanner, and Reflection. - 📝 **Tasks** define problems to solve, complete with scoring functions. Tasks can be chained into TaskGroups for complex multi-stage challenges. *Corral* currently ships **8 environments**, **97 tools**, **115 tasks**, and **786 subtasks** spanning chemistry, physics, and materials science. ### 🌍 Environments | Environment | Description | 🔧 Tools | 📝 Tasks/scope | 🔭 Scopes | ⏱️ Avg. trace length | |---|---|:---:|:---:|:---:|:---:| | 🧫 **Inorganic Qualitative Analysis** | Identify unknown cations in solution through systematic wet-lab procedures (reagent addition, flame tests, pH measurement, centrifugation, etc.). Observations are computed from thermodynamic data. Three scopes progressively increase the number of candidate ions. | 14 | 10 | 3 | 39.4 | | ⚡ **Circuit Inference** | Recover the topology and component values of a hidden resistor network from pairwise resistance measurements. Tools provide series/parallel calculations, delta-wye transforms, and circuit validation. | 9 | 6 | 1 | 15.0 | | 🔭 **Spectroscopic Structure Elucidation** | Determine the molecular structure of an unknown compound by requesting and interpreting spectroscopic data (MS, NMR, HSQC, IR) alongside reference databases for chemical shifts and isotope distributions. | 16 | 20 | 2 | 15.1 | | 🧬 **Retrosynthetic Planning** | Design multi-step synthetic routes to target molecules under cost, step-count, and commercial-availability constraints, using a template catalogue and functional-group detection tools. | 15 | 8 | 3 | 25.5 | | 🤖 **ML-based Property Prediction** | Assemble a complete ML pipeline to predict formation energies of material polymorphs using data from the Materials Project, covering feature engineering, XGBoost training, and cross-validation. | 14 | 3 | 1 | 16.6 | | 🔬 **AFM Experiment Execution** | Analyze and interpret atomic force microscopy data for nanoscale surface characterization, including topographical and mechanical property measurements. | 6 | 1 | 4 | 26.3 | | ⚛️ **Molecular Simulation** | Design and execute molecular dynamics simulations with LAMMPS to predict materials properties, covering the full workflow from crystal structure retrieval to force-field queries and log analysis. | 8 | 2–3 | 2 | 30.4 | | 🏗️ **Adsorption Surface Construction** | Build adsorbate–slab configurations from bulk crystal structures for heterogeneous catalysis studies, integrating Materials Project retrieval, slab generation, and adsorption-site enumeration. | 15 | 3 | 1 | 19.6 | --- ## 🗂️ Dataset Structure ### Configs The dataset exposes one config per annotated model: `claude_sonnet_45`, `gpt_4o`, and `gpt_oss_120b`. ### Data Splits The dataset exposes a single `train` split. ### Data Instances Each row corresponds to one **annotated trace file**. These rows contain the trace together with its **LLM epistemic annotations** and reflect cases where the annotator judged that the agent did not reason scientifically. --- ## 🏗️ Dataset Creation ### Curation Rationale This dataset was created as part of *Corral* to support targeted inspection of **scientific reasoning failures** beyond end-task success. By releasing trace annotations generated by an automatic annotator, it provides a focused resource for analyzing epistemic failure patterns. ### Source Data The traces were derived from *Corral* evaluation runs across environments and models. A downstream **LLM annotator** labeled the traces with epistemic annotations and identified cases suggesting that the agent did not reason scientifically. Each retained row corresponds to one annotated file. --- ## 🔗 Relation to Other Corral Artifacts This dataset is one component of the broader *Corral* release and is best interpreted together with the matching task definitions, execution traces, reports, aggregate results, and reasoning annotations available in the [*Corral* collection](https://huggingface.co/collections/jablonkagroup/corral). --- ## 📄 Citation ```bibtex @article{ríos-garcía2026ai, title = {AI scientists produce results without reasoning scientifically}, author = {Martiño Ríos-García and Nawaf Alampara and Chandan Gupta and Indrajeet Mandal and Sajid Mannan and Ali Asghar Aghajani and N. M. Anoop Krishnan and Kevin Maik Jablonka}, year = {2026}, journal = {arXiv preprint arXiv: 2604.18805} } ``` ## 📜 License This dataset is released under the [MIT License](https://opensource.org/licenses/MIT). ## Changelog ### 2026-04-22 - Initial release of the dataset card.