MARICL uses multi-agent LLMs to derive correction formulas from base-model residuals and validates mechanistic generalization by freezing and applying those formulas to held-out batches in a protein synthesis dataset.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
From Residuals to Reasons: LLM-Guided Mechanism Inference from Tabular Data
MARICL uses multi-agent LLMs to derive correction formulas from base-model residuals and validates mechanistic generalization by freezing and applying those formulas to held-out batches in a protein synthesis dataset.