aiqbench analysis
This study challenges the assumption that structural differences in discovered circuits indicate distinct underlying mechanisms, revealing instead a phenomenon of "phantom specialization" where varying input frequencies yields structurally unique but functionally equivalent subgraphs. The research demonstrates that standard evaluation practices often obscure this many-to-one mapping between structure and function, suggesting that edge-level analysis is necessary to accurately interpret model behaviors. Consequently, researchers should exercise caution when inferring mechanistic diversity solely from the topology of extracted circuits without verifying functional interchangeability across conditions.
Many Circuits, One Mechanism: Input Variation and Evaluation Granularity in Circuit Discovery
arXiv:2606.06267v2 Announce Type: replace Abstract: Circuit discovery methods identify subgraphs that explain model behaviors, and structural differences between discovered circuits are commonly interpreted as evidence of distinct mechanisms. We test this assumption by varying input-token frequency while holding the task fixed. The discovered circuits appear specialized by frequency when compared structurally, but functional and representational analyses show no reliable evidence of corresponding differences. We term this mismatch phantom specialization. Using the Literal Sequence Copying task across four frequency bands plus a frequency-weighted control, we extract 75 circuits from five Pythia models (70M-1.4B). We find that structurally distinct circuits implement the same computation: band-specific edges transfer broadly across bands, a core shared across most bands recovers at least 99% of circuit performance in models above 70M, and causal interchange interventions confirm that internal representations are interchangeable across frequency bands. A smaller subject-verb agreement replication shows the same pattern: circuits differ structurally but transfer broadly across bands, and the core shared by most bands recovers nearly all circuit accuracy. Repeated extractions within the same band further suggest that discovery algorithms sample from an equivalence class of valid subgraphs rather than recovering a unique mechanism. Standard evaluation practice obscures this pattern: source-level evaluation inflates apparent faithfulness, while edge-level evaluation reveals the many-to-one mapping from structure to function. We do not test specialization at the level of token positions or features inside a component. Our results show that structural differences between circuits are not sufficient evidence for distinct mechanisms, and that exposing this requires edge-level evaluation and cross-condition transfer tests.
Abstract:Circuit discovery methods identify subgraphs that explain model behaviors, and structural differences between discovered circuits are commonly interpreted as evidence of distinct mechanisms. We test this assumption by varying input-token frequency while holding the task fixed. The discovered circuits appear specialized by frequency when compared structurally, but functional and representational analyses show no reliable evidence of corresponding differences. We term this mismatch phantom specialization. Using the Literal Sequence Copying task across four frequency bands plus a frequency-weighted control, we extract 75 circuits from five Pythia models (70M-1.4B). We find that structurally distinct circuits implement the same computation: band-specific edges transfer broadly across bands, a core shared across most bands recovers at least 99% of circuit performance in models above 70M, and causal interchange interventions confirm that internal representations are interchangeable across frequency bands. A smaller subject-verb agreement replication shows the same pattern: circuits differ structurally but transfer broadly across bands, and the core shared by most bands recovers nearly all circuit accuracy. Repeated extractions within the same band further suggest that discovery algorithms sample from an equivalence class of valid subgraphs rather than recovering a unique mechanism. Standard evaluation practice obscures this pattern: source-level evaluation inflates apparent faithfulness, while edge-level evaluation reveals the many-to-one mapping from structure to function. We do not test specialization at the level of token positions or features inside a component. Our results show that structural differences between circuits are not sufficient evidence for distinct mechanisms, and that exposing this requires edge-level evaluation and cross-condition transfer tests.
Computation and Language (cs.CL)
arXiv:2606.06267 [cs.CL]
(or arXiv:2606.06267v2 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2606.06267 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 344 words
Related coverage
Many Circuits, One Mechanism: Input Variation and Evaluation Granularity in Circuit Discovery
arXiv · Jun 5, 2026
J++ Lens: Jacobian Filtering Enables More Faithful Workspace Lenses
LessWrong · Oct 8, 2026
Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible
LessWrong · Oct 5, 2026
Lattice-Based Cryptography Explained
Quantum Zeitgeist · Aug 15, 2026