# Who Gets Credit or Blame? Attributing Accountability in Modern AI Systems

Source: [Who Gets Credit or Blame? Attributing Accountability in Modern AI Systems](<https://www.sota2.com/research/paper/68fa7b211e209131cea86ba5>)

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

Data updated: 2026-06-02T05:00:05.013Z

Authors: Shichang Zhang, Hongzhe Du, Jiaqi W. Ma, Himabindu Lakkaraju

Published: 2025-05-30T19:27:39.000Z

## Abstract

Modern AI systems are typically developed through multiple stages-pretraining, fine-tuning rounds, and subsequent adaptation or alignment, where each stage builds on the previous ones and updates the model in distinct ways. This raises a critical question of accountability: when a deployed model succeeds or fails, which stage is responsible, and to what extent? We pose the accountability attribution problem for tracing model behavior back to specific stages of the model development process. To address this challenge, we propose a general framework that answers counterfactual questions about stage effects: how would the model's behavior have changed if the updates from a particular stage had not occurred? Within this framework, we introduce estimators that efficiently quantify stage effects without retraining the model, accounting for both the data and key aspects of model optimization dynamics, including learning rate schedules, momentum, and weight decay. We demonstrate that our approach successfully quantifies the accountability of each stage to the model's behavior. Based on the attribution results, our method can identify and remove spurious correlations learned during image classification and text toxicity detection tasks that were developed across multiple stages. Our approach provides a practical tool for model analysis and represents a significant step toward more accountable AI development.

## 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.

- [Accountability Attribution on CelebA (test)](<https://www.sota2.com/research/sota/accountability-attribution-on-celeba-test.md>): Accountability Attribution; CelebA (test); Stage 1 Score: 55.24.

- [Accountability Attribution on CivilComments (test)](<https://www.sota2.com/research/sota/accountability-attribution-on-civilcomments-test.md>): Accountability Attribution; CivilComments (test); Stage 1 Score: 85.76.

- [Correlation of Stage Effects on MNIST (test)](<https://www.sota2.com/research/sota/correlation-of-stage-effects-on-mnist-test.md>): Correlation of Stage Effects; MNIST (test); Insert Correlation: 94.44.

- [Accountability Attribution on X-Ray (test)](<https://www.sota2.com/research/sota/accountability-attribution-on-x-ray-test.md>): Accountability Attribution; X-ray (test); Stage 1 Score: 93.26.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2506.00175>)

[PDF](<https://arxiv.org/pdf/2506.00175.pdf>)
