Operator-ready prompt for reuse, tuning, and workspace runs.
This item is set up for developers who want to inspect the original language, fork it into Workspace, and adapt the evidence model without losing the source prompt structure.
Implementation handoffs, eval setup, and prompt tuning where you need the original structure intact.
Inspect first, copy once, then fork into Workspace when you want variants, notes, and model settings attached to the same run.
Swap domain facts, examples, and any hard-coded entities for your own context.
Tighten the evidence or verification requirement if this is headed toward production.
Decide which failure mode you want to evaluate first before you branch the prompt.
This prompt already carries implementation detail, tool context, and a final-output instruction. Keep that structure intact when you tune it, or your comparison runs get noisy fast.
Open this prompt inside Workspace when you want a live iteration loop.
Copy for quick reuse, or run it in Workspace to keep prompt variants, model settings, and prompt-history changes in one place.
Structured source with 1 active lines to adapt.
Already linked to a challenge workflow.
Sign in to keep private prompt variations.
Prompt content
Original prompt text with formatting preserved for inspection and clean copy.
Select 3-5 challenging scientific problems (e.g., from a simulated FrontierScience-like benchmark) and run your complete LangGraph-based agent system through them. Document the system's final conclusions, but critically, also trace and present the entire reasoning path, including tool calls, intermediate thoughts, and any self-correction steps. Evaluate the transparency and verifiability of the process.
Adaptation plan
Keep the source stable, then branch your edits in a predictable order so the next prompt run is easier to evaluate.
Preserve the rubric, target behavior, and pass-fail criteria as the baseline for evaluation.
Adjust fixtures, mocks, and thresholds to the system under test instead of weakening the assertions.
Make sure the prompt catches regressions instead of just mirroring the happy-path examples.
Copy once for a pristine source snapshot, then move the prompt into Workspace when you want variants, run history, and side-by-side tuning without losing the original.
Prompt diagnostics
Quick signals for how structured this prompt already is and where adaptation work is likely to happen first.
This prompt is mostly narrative and instruction-driven, so you can adapt examples and output constraints first without disturbing the structure.
Graph-Based Scientific Reasoning Agent
Inspired by OpenAI's FrontierScience benchmark, this challenge focuses on developing an advanced agent system capable of tackling expert-level scientific reasoning problems. Participants will design and implement a graph-based workflow that simulates the scientific method – from hypothesis generation and experimental design (simulated) to data analysis and conclusion formulation. The system will leverage state-of-the-art LLMs for complex problem-solving and incorporate MCP-enabled tools for integrating with scientific databases and symbolic computation engines. The emphasis is on building a robust, verifiable reasoning pipeline that can explain its steps and adapt its approach based on intermediate results, showcasing extended thinking capabilities and hybrid reasoning. The core of this challenge involves using LangGraph to define a Directed Acyclic Graph (DAG) that represents stages of scientific inquiry. Agents, powered by GPT-5.2 and potentially DeepSeek-V3 for specialized tasks, will interact within this graph, using DSPy for optimizing prompts to achieve scientific accuracy and minimize hallucinations. Developers will integrate MCP tools for accessing external knowledge (e.g., ArXiv, PubMed) and computational resources, and implement adaptive thinking budgets to allow for deeper analysis on critical scientific junctures. The final system should not only solve problems but also explain its reasoning process transparently.
Use the challenge page to recover the original task boundaries before you tune the prompt. That keeps your variants grounded in the same evaluation target instead of drifting into a different problem.