Visualizing Graph-to-Answer Mechanism Recovery in Materials-Science Hypothesis Generation

可视化材料科学假设生成中的“图到答案”机制恢复

Abstract: AI co-scientists can generate fluent materials-science hypotheses, but fluency does not show that an answer preserves a scientifically meaningful mechanism. We present a graph-to-answer mechanism-tracing case study for Graph-PRefLexOR-8B, a Qwen3-8B model adapted to expose distinct stages for brainstorming, graph construction, pattern extraction, and synthesis.

摘要： 人工智能科研助手能够生成流畅的材料科学假设，但流畅性并不代表答案保留了具有科学意义的机制。我们针对 Graph-PRefLexOR-8B 模型提出了一个“图到答案”的机制追踪案例研究。该模型基于 Qwen3-8B 适配而成，旨在展示头脑风暴、图构建、模式提取和综合分析等不同阶段。

We organize semantic backtracking, graph corruption, activation-based recovery measurements, and layer-by-token-region grids into a visual diagnostic workflow for inspecting this pathway. Across 100 open-ended materials-science questions, final answers remain closest to the model’s own structured stages, especially synthesis.

我们将语义回溯、图破坏、基于激活的恢复测量以及层级-标记区域网格组织成一个可视化诊断工作流，用于检查这一路径。在 100 个开放式材料科学问题的测试中，最终答案与模型自身的结构化阶段（尤其是综合分析阶段）保持最紧密的联系。

Under graph corruption, a full sweep over 37 residual-stream checkpoints, the embedding output and 36 transformer blocks, shows little mechanism recovery in the earlier transition region at layers 7—10, recovery instead concentrates in late synthesis and answer-start regions around layers 30 and 36.

在图破坏测试中，对 37 个残差流检查点（包括嵌入输出和 36 个 Transformer 块）进行全面扫描后发现，在第 7 至 10 层的早期过渡区域几乎没有机制恢复；相反，恢复主要集中在第 30 层和第 36 层附近的后期综合分析和答案起始区域。

The workflow is intended to help scientists and model developers identify where a generated hypothesis loses or regains mechanism support before it is passed to downstream experimental planning.

该工作流旨在帮助科学家和模型开发者在生成的假设被传递到下游实验规划之前，识别出其在何处丢失或重新获得了机制支持。