BibTeX

@inProceedings{farahani-etal-2026-copy-366988,
	title        = {To Copy or Not to Copy: Copying Is Easier to Induce Than Recall
},
	abstract     = {Language models used in retrieval-augmented settings choose between parametric knowledge stored in their weights and contextual information in the prompt. This work presents a mechanistic study of that choice: using an approach based on steering vectors, we carry out interventions at selected layers and token spans, and we are able to toggle the model’s behavior in two directions: Copy→Recall (suppressing context use) and Recall→Copy (inducing the model to copy any token from the context). Experiments on three architectures (decoder-only and encoder/decoder) and two open-domain QA benchmarks show consistent behavior shifts under moderate scaling while monitoring accuracy and fluency. Mechanistic analyses of attention routing, MLP contributions, and layer-wise probability trajectories reveal an asymmetry: inducing copying is an easy "reactivation" process that can be triggered at different locations in the input, while restoring recall is a "suppression" process that is more fragile and strongly tied to object-token interventions.},
	booktitle    = {Findings of the Association for Computational Linguistics: EMNLP 2026, Budapest, Hungary},
	author       = {FARAHANI, MEHRDAD and Penzkofer, Franziska and Johansson, Richard},
	year         = {2026},
	publisher    = {Association for Computational Linguistics},
	ISBN         = {979-8-89176-462-0},
	pages        = {50027–50090},
}