# Towards Unifying Interpretability and Control: Evaluation via Intervention

Source: [Towards Unifying Interpretability and Control: Evaluation via Intervention](<https://www.sota2.com/research/paper/68fa76e51e209131cea6fbea>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-28T13:57:13.387Z

Authors: Usha Bhalla, Suraj Srinivas, Asma Ghandeharioun, Himabindu Lakkaraju

Published: 2024-11-07T04:52:18.000Z

## Abstract

With the growing complexity and capability of large language models, a need to understand model reasoning has emerged, often motivated by an underlying goal of controlling and aligning models. While numerous interpretability and steering methods have been proposed as solutions, they are typically designed either for understanding or for control, seldom addressing both. Additionally, the lack of standardized applications, motivations, and evaluation metrics makes it difficult to assess methods' practical utility and efficacy. To address the aforementioned issues, we argue that intervention is a fundamental goal of interpretability and introduce success criteria to evaluate how well methods can control model behavior through interventions. To evaluate existing methods for this ability, we unify and extend four popular interpretability methods-sparse autoencoders, logit lens, tuned lens, and probing-into an abstract encoder-decoder framework, enabling interventions on interpretable features that can be mapped back to latent representations to control model outputs. We introduce two new evaluation metrics: intervention success rate and coherence-intervention tradeoff, designed to measure the accuracy of explanations and their utility in controlling model behavior. Our findings reveal that (1) while current methods allow for intervention, their effectiveness is inconsistent across features and models, (2) lens-based methods outperform SAEs and probes in achieving simple, concrete interventions, and (3) mechanistic interventions often compromise model coherence, underperforming simpler alternatives, such as prompting, and highlighting a critical shortcoming of current interpretability approaches in applications requiring control.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Addition on 12-digit addition 3 terms](<https://www.sota2.com/research/sota/addition-on-12-digit-addition-3-terms.md>): Addition; 12-digit addition 3 terms; Token Accuracy: 94.7.

- [Token-level Addition on Addition T3 (val)](<https://www.sota2.com/research/sota/token-level-addition-on-addition-t3-val.md>): Token-level Addition; Addition T3 (val); Token Accuracy: 88.27.

- [Addition on 10-digit addition 3 terms](<https://www.sota2.com/research/sota/addition-on-10-digit-addition-3-terms.md>): Addition; 10-digit addition 3 terms; Token Accuracy: 88.27.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2411.04430>)

[PDF](<https://arxiv.org/pdf/2411.04430.pdf>)
