# CIRCUS: Circuit Consensus under Uncertainty via Stability Ensembles

arXiv: https://arxiv.org/abs/2603.00523 · HTML: https://mlsanity.com/paper/2603.00523/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Circuits and mechanistic analysis, one of 13 directions in Mechanistic interpretability.
- Builds on 14 of its 19 references that are in the map.
- First posted on arXiv 28 Feb 2026.

## Placed in

- Mechanistic interpretability / [Circuits and mechanistic analysis](https://mlsanity.com/mechinterp/circuits-and-mechanistic-analysis/)

## Builds on (most cited first, 14 of 14)

- [Towards A Rigorous Science of Interpretable Machine Learning](https://mlsanity.com/paper/1702.08608/) (2017)
- [Towards Best Practices of Activation Patching in Language Models: Metrics and Methods](https://mlsanity.com/paper/2309.16042/) (2023)
- [Open Problems in Mechanistic Interpretability](https://mlsanity.com/paper/2501.16496/) (2025)
- [A Practical Review of Mechanistic Interpretability for Transformer-Based Language Models](https://mlsanity.com/paper/2407.02646/) (2024)
- [Rethinking Stability for Attribution-based Explanations](https://mlsanity.com/paper/2203.06877/) (2022)
- [Explaining Predictive Uncertainty by Exposing Second-Order Effects](https://mlsanity.com/paper/2401.17441/) (2024)
- [My Teacher Thinks The World Is Flat! Interpreting Automatic Essay Scoring Mechanism](https://mlsanity.com/paper/2012.13872/) (2020)
- [AES Systems Are Both Overstable And Oversensitive: Explaining Why And Proposing Defenses](https://mlsanity.com/paper/2109.11728/) (2021)
- [MINIMAL: Mining Models for Data Free Universal Adversarial Triggers](https://mlsanity.com/paper/2109.12406/) (2021)
- [Ensembling improves stability and power of feature selection for deep learning models](https://mlsanity.com/paper/2210.00604/) (2022)
- [Ensembling Sparse Autoencoders](https://mlsanity.com/paper/2505.16077/) (2025)
- [Robust Explanations Through Uncertainty Decomposition: A Path to Trustworthier AI](https://mlsanity.com/paper/2507.12913/) (2025)
- [Measuring Uncertainty in Transformer Circuits with Effective Information Consistency](https://mlsanity.com/paper/2509.07149/) (2025)
- [Aggregate Models, Not Explanations: Improving Feature Importance Estimation](https://mlsanity.com/paper/2602.11760/) (2026)

## Built on by (most cited first, 1 of 1)

- [Circuit Condensation: Post-Training that Concentrates a Behavior's Causal Circuit](https://mlsanity.com/paper/2608.27254/) (2026)

## Abstract (arXiv)

Every mechanistic circuit carries an invisible asterisk: it reflects not just the model's computation, but the analyst's choice of pruning threshold. Change that choice and the circuit changes, yet current practice treats a single pruned subgraph as ground truth with no way to distinguish robust structure from threshold artifacts. We introduce CIRCUS, which reframes circuit discovery as a problem of uncertainty over explanations. CIRCUS prunes one attribution graph under B configurations, assigns each edge an empirical inclusion frequency s(e) in [0,1] measuring how robustly it survives across the configuration family, and extracts a consensus circuit of edges present in every view. This yields a principled core/contingent/noise decomposition (analogous to posterior model-inclusion indicators in Bayesian variable selection) that separates robust structure from threshold-sensitive artifacts, with negligible overhead. On Gemma-2-2B and Llama-3.2-1B, consensus circuits are 40x smaller than the union of all configurations while retaining comparable influence-flow explanatory power, consistently outperform influence-ranked and random baselines, and are confirmed causally relevant by activation patching.