Abstract:Multimodal LLMs are increasingly used to assist scientific peer review, where a core requirement is verifying whether claims in a paper are supported by its evidence. Prior work has shown that models perform substantially better at this task when the evidence is a table than when it is a chart of the same underlying data. This raises the question of whether models fail to extract information from charts, or do they extract it but fail to use it when forming their prediction? We study this question through layer-wise linear probing and attention analysis on three open-weight VLMs over table and chart evidence, representing the same underlying data. We find consistent evidence for the latter. Chart information is encoded in the models' intermediate representations but does not reach the prediction position, a gap that is absent for tables and holds across all conditions tested. Attention analysis further reveals that this disconnect takes two architecturally distinct forms across model families. These findings reframe the table-chart gap as a failure of how encoded visual information is routed at prediction time, rather than a failure of encoding itself.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2606.01679 [cs.CL] |
| (or arXiv:2606.01679v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2606.01679 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sunisth Kumar [view email]
[v1]
Mon, 1 Jun 2026 04:39:13 UTC (264 KB)