aiqbench analysis
This research addresses a critical deployment risk where standard model quantization preserves predictive accuracy but degrades the stability of algorithmic recourse actions. The authors introduce Counterfactual-Faithful Quantization (CFQ), a method that jointly optimizes bit allocation to maintain both prediction fidelity and the validity of actionable interventions required for decision reversal. By validating on high-stakes datasets like COMPAS, this work establishes new metrics—Validity Drop and Recourse Gap—to ensure compressed models remain reliable in real-world causal inference systems.
When Bits Break Recourse: Counterfactual-Faithful Quantization
arXiv:2605.17160v2 Announce Type: replace Abstract: Model quantization is widely used to reduce memory, latency, and deployment cost, and is typically judged by whether predictive accuracy is preserved. In decision systems that provide algorithmic recourse, however, accuracy preservation is not sufficient: a small actionable change that flips the decision of a full-precision model may fail after quantization, or require a substantially larger intervention. This paper studies this deployment mismatch and introduces counterfactual sensitivity under quantization, a framework for measuring how compression changes recourse behavior. We propose two metrics: Validity Drop (VD), which measures the fraction of full-precision recourse actions that no longer achieve the target outcome after quantization, and Counterfactual Recourse Gap (CRG), which measures the increase in minimal recourse cost under the quantized model. To mitigate this failure mode, we introduce Counterfactual-Faithful Quantization (CFQ), a quantization-aware training method that jointly learns quantizer parameters and mixed-precision bit allocation while preserving the target prediction at teacher-generated recourse points. CFQ is compatible with standard LSQ/PACT-style quantizers and mixed-precision policies, and can also be instantiated as a training-free calibration procedure for post-training quantization. Experiments on Adult, German Credit, and COMPAS show that standard QAT and mixed-precision baselines can preserve accuracy while substantially degrading recourse stability. At matched accuracy and bit budget, CFQ consistently reduces VD and CRG; for example, on Adult, CFQ reduces VD/CRG from $0.121/0.162$ for an accuracy-centric mixed-precision baseline to $0.061/0.071$.
Abstract:Model quantization is widely used to reduce memory, latency, and deployment cost, and is typically judged by whether predictive accuracy is preserved. In decision systems that provide algorithmic recourse, however, accuracy preservation is not sufficient: a small actionable change that flips the decision of a full-precision model may fail after quantization, or require a substantially larger intervention. This paper studies this deployment mismatch and introduces counterfactual sensitivity under quantization, a framework for measuring how compression changes recourse behavior. We propose two metrics: Validity Drop (VD), which measures the fraction of full-precision recourse actions that no longer achieve the target outcome after quantization, and Counterfactual Recourse Gap (CRG), which measures the increase in minimal recourse cost under the quantized model. To mitigate this failure mode, we introduce Counterfactual-Faithful Quantization (CFQ), a quantization-aware training method that jointly learns quantizer parameters and mixed-precision bit allocation while preserving the target prediction at teacher-generated recourse points. CFQ is compatible with standard LSQ/PACT-style quantizers and mixed-precision policies, and can also be instantiated as a training-free calibration procedure for post-training quantization. Experiments on Adult, German Credit, and COMPAS show that standard QAT and mixed-precision baselines can preserve accuracy while substantially degrading recourse stability. At matched accuracy and bit budget, CFQ consistently reduces VD and CRG; for example, on Adult, CFQ reduces VD/CRG from $0.121/0.162$ for an accuracy-centric mixed-precision baseline to $0.061/0.071$.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2605.17160 [cs.LG]
(or arXiv:2605.17160v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2605.17160 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 · 325 words
Related coverage
MosaicQuant: Inlier-Outlier Disaggregation for Unified 4-Bit LLM Quantization
arXiv · Jun 17, 2026
Can Large Language Models Still Explain Themselves? Investigating the Impact of Quantization on Self-Explanations
arXiv · Aug 26, 2026
[AINews] not much happened today
Latent Space · Oct 3, 2026
Quantization and Pruning Methods to Make Your LLM Leaner
KDnuggets · Aug 28, 2026