{"id":"arxiv:2407.14008","kind":"preprint","title":"Investigating the Indirect Object Identification circuit in Mamba","authors":["Danielle Ensign","Adrià Garriga-Alonso"],"author_count":2,"year":2024,"published":"2024-07-19","venue":null,"cited_by":null,"doi":null,"arxiv_id":"2407.14008","nct_id":null,"source_url":"https://arxiv.org/abs/2407.14008","archive_url":"https://space-kitty.com/papers/arxiv:2407.14008","updated":"2024-07-22","pmid":null,"pmcid":null,"openalex_id":null,"pdf_url":"https://arxiv.org/pdf/2407.14008","abstract":"How well will current interpretability techniques generalize to future models? A relevant case study is Mamba, a recent recurrent architecture with scaling comparable to Transformers. We adapt pre-Mamba techniques to Mamba and partially reverse-engineer the circuit responsible for the Indirect Object Identification (IOI) task. Our techniques provide evidence that 1) Layer 39 is a key bottleneck, 2) Convolutions in layer 39 shift names one position forward, and 3) The name entities are stored linearly in Layer 39's SSM. Finally, we adapt an automatic circuit discovery tool, positional Edge Attribution Patching, to identify a Mamba IOI circuit. Our contributions provide initial evidence that circuit-based mechanistic interpretability tools work well for the Mamba architecture.","abstract_license":"CC0-1.0 (arXiv descriptive metadata)","license":null,"is_open_access":false,"retracted":false,"peer_reviewed":false,"categories":["cs.LG"],"details":null,"provenance":"arxiv","attribution":"Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.","topics":[],"cited_in_dossiers":[]}