Search papers, labs, and topics across Lattice.
This study investigates how the open-weight google/gemma-4-E4B-it model encodes and utilizes materials science mechanisms, revealing that information is represented in three distinct forms: readable concepts in hidden states, transformations between states, and causal control over engineering outputs. By employing matched direct and Jacobian vocabulary readouts alongside a counterfactual benchmark, the authors demonstrate that the model can effectively reproduce concept ranks and identify mechanism families in held-out materials descriptions. The findings highlight that physical relationships are more discernible through controlled state transformations than through absolute states, suggesting a nuanced understanding of how language models engage with scientific concepts.
The model reveals that physical relationships in materials science are more accessible through controlled transformations than through static representations, challenging assumptions about how LLMs encode scientific knowledge.
Large language models can answer scientific questions, yet a correct output does not reveal whether the model represents or uses the governing physics. Here we show that materials science mechanism information in the open-weight google/gemma-4-E4B-it model has three experimentally separable forms: concepts are readable in individual hidden states, constitutive orientation is carried by controlled transformations between states, and selected internal representations causally control engineering answers. We combine matched direct and Jacobian vocabulary readouts, option-free state geometry, a 60-law counterfactual benchmark and causal interventions. In 50 held-out materials descriptions, three independently fitted Jacobian lenses reproduced concept ranks, and target-free word sets from both readouts enabled blinded identification of 9 of 10 mechanism families. A separate 72-prompt benchmark produced mechanism-specific hidden-state neighborhoods, but an exact graph audit showed that this apparent physical organization was equally explained by numerical comparison. We therefore compared otherwise identical prompts in which only the direction of the physical input was reversed, asking whether the resulting hidden-state movement followed the supplied constitutive law. These state transformations ordered direct, physically neutral and inverse laws across 60 frozen relations and correctly oriented 39 of 40 directional laws, whereas lexical controls were near chance. Bidirectional interventions shifted answer probabilities toward or away from the physically appropriate outcome across all 12 matched cases, while counterfactual state patches transferred opposing decision signals across mechanisms and answer formats. Physical relationships were therefore more visible in controlled state changes than in absolute states alone.
