Certified Interventional Fidelity: Anytime-Valid, Adaptive Evaluation of Causal Claims in Mechanistic Interpretability

• URL: http://arxiv.org/abs/2607.08349v1
• Date: Thu, 09 Jul 2026 10:55:04 GMT
• Title: Certified Interventional Fidelity: Anytime-Valid, Adaptive Evaluation of Causal Claims in Mechanistic Interpretability
• Abstract summary: Certified Interventional Fidelity (CIF) is a statistical layer for interventional interpretability evaluations.<n>CIF certifies high-fidelity claims, shows when apparent method differences are not statistically supported, and makes sensitivity to the intervention distribution explicit.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Mechanistic interpretability often evaluates explanations by intervening on a model: swapping hidden states, patching activations, ablating components, or comparing a compressed model to the original one. These experiments are usually summarized by a point estimate, even though the evaluation may be monitored while it runs or adapted toward suspected failures. This makes it hard to tell whether a reported fidelity or patching effect is a stable causal claim or a consequence of finite sampling and evaluation choices. We introduce Certified Interventional Fidelity (CIF), a statistical layer for interventional interpretability evaluations. CIF first writes the quantity being reported as a causal estimand: an expectation of a bounded score over a stated input distribution and a stated intervention distribution. It then provides confidence intervals and anytime-valid confidence sequences for this estimand, including under adaptive intervention sampling via bounded mixture importance weighting. We instantiate CIF with Hoeffding-style sequences and variance-adaptive betting sequences, the latter reducing certification cost by 10-30x in our experiments. On MNIST abstractions and GPT-2 Small IOI circuits, CIF certifies high-fidelity claims, shows when apparent method differences are not statistically supported, and makes sensitivity to the intervention distribution explicit.

Related papers

• Diffusion-Guided Search via Exponential Tilting (DiffTilt): An Application to Falsification of Safety-Critical Systems [3.963282613048148]
  Existing falsification approaches rely on conditional sampling strategies that factor the joint distribution over environments and system executions.<n>This paper develops DiffTilt, a distributional framework that exponentially tilts a diffusion model-induced joint distribution over environments and executions.
  arXiv  Detail & Related papers  (2026-07-25T10:28:47Z)
• Accelerating A/B-Tests with Counterfactual Estimation: Reducing Variance through Policy Overlap [8.291484471359633]
  We identify a critical inefficiency in the standard A/B-testing protocol.<n>When a treatment and control policy agree on an action, the resulting outcome contributes noise but no signal regarding the treatment effect.<n>We propose a novel experimental protocol that exploits this policy overlap to accelerate experimentation.
  arXiv  Detail & Related papers  (2026-07-16T06:08:52Z)
• Heuristic Pathologies and Further Variance Reduction via Uncertainty Propagation in the AIVAT Family of Techniques [62.94836578759359]
  We show how the uncertainty can be propagated to quantify the uncertainty of AIVAT estimates.<n>In our experiments, we use a dataset of 10,000 poker hands to demonstrate our pathology and uncertainty results.
  arXiv  Detail & Related papers  (2026-05-14T02:04:26Z)
• FACTOR: Counterfactual Training-Free Test-Time Adaptation for Open-Vocabulary Object Detection [63.91351553178842]
  FACTOR is a training-free Test-time adaptation for Open-vocabulary object detection.<n>It quantifies attribute sensitivity, semantic relevance, and prediction variation to selectively suppress attribute-dependent predictions.<n>Experiments on PASCAL-C, COCO-C, and FoggyCityscapes show that FACTOR consistently outperforms prior TTA methods.
  arXiv  Detail & Related papers  (2026-05-05T02:31:18Z)
• FACT-E: Causality-Inspired Evaluation for Trustworthy Chain-of-Thought Reasoning [49.65751420291115]
  Chain-of-Thought (CoT) prompting has improved LLM reasoning, but models often generate explanations that appear coherent while containing unfaithful intermediate steps.<n>We propose FACT-E, a causality-inspired framework for evaluating CoT quality.<n>We show that FACT-E improves reasoning-trajectory selection and yields stronger in-context learnings.
  arXiv  Detail & Related papers  (2026-04-12T15:35:08Z)
• SPAN: Continuous Modeling of Suspicion Progression for Temporal Intention Localization [26.07264704956791]
  We propose the Suspicion Progression Analysis Network (SPAN), which shifts from discrete classification to continuous regression.<n>SPAN achieves a 2.74% mAP gain in low-frequency cases, demonstrating its superior ability to capture subtle behavioral changes.
  arXiv  Detail & Related papers  (2025-10-23T04:20:07Z)
• Adaptive Individual Uncertainty under Out-Of-Distribution Shift with Expert-Routed Conformal Prediction [8.306260695214972]
  We introduce Trustworthy Expert Split-conformal with Scaled Estimation for Efficient Reliable Adaptive intervals (TESSERA)<n>TESSERA provides per-sample uncertainty with reliable coverage guarantee, informative and adaptive prediction interval widths that track the absolute error.<n>We evaluate on protein-ligand binding affinity prediction under both independent and identically distributed (i.i.d.) and scaffold-based out-of-distribution (OOD) splits.
  arXiv  Detail & Related papers  (2025-10-17T01:51:33Z)
• Proximity-Based Evidence Retrieval for Uncertainty-Aware Neural Networks [6.9681910774977815]
  This work proposes an evidence-retrieval mechanism for uncertainty-aware decision-making.<n>For each test instance, exemplars are retrieved in an embedding space; their predictive distributions are fused via Dempster-Shafer theory.<n>Because the supporting evidences are explicit, decisions are transparent and auditable.
  arXiv  Detail & Related papers  (2025-09-11T13:12:22Z)
• TrustLoRA: Low-Rank Adaptation for Failure Detection under Out-of-distribution Data [62.22804234013273]
  We propose a simple failure detection framework to unify and facilitate classification with rejection under both covariate and semantic shifts.<n>Our key insight is that by separating and consolidating failure-specific reliability knowledge with low-rank adapters, we can enhance the failure detection ability effectively and flexibly.
  arXiv  Detail & Related papers  (2025-04-20T09:20:55Z)
• ACTIVA: Amortized Causal Effect Estimation via Transformer-based Variational Autoencoder [7.987204219322316]
  We propose ACTIVA, a conditional variational autoencoder architecture for amortized causal inference.<n>ACTIVA learns a latent representation conditioned on observational inputs and intervention queries, enabling zero-shot inference.<n>We provide theoretical insights showing that ACTIVA predicts interventional distributions as mixtures over observationally equivalent causal models.
  arXiv  Detail & Related papers  (2025-03-03T08:28:25Z)
• Error-quantified Conformal Inference for Time Series [55.11926160774831]
  Uncertainty quantification in time series prediction is challenging due to the temporal dependence and distribution shift on sequential data.<n>We propose itError-quantified Conformal Inference (ECI) by smoothing the quantile loss function.<n>ECI can achieve valid miscoverage control and output tighter prediction sets than other baselines.
  arXiv  Detail & Related papers  (2025-02-02T15:02:36Z)
• Mitigating LLM Hallucinations via Conformal Abstention [70.83870602967625]
  We develop a principled procedure for determining when a large language model should abstain from responding in a general domain. We leverage conformal prediction techniques to develop an abstention procedure that benefits from rigorous theoretical guarantees on the hallucination rate (error rate) Experimentally, our resulting conformal abstention method reliably bounds the hallucination rate on various closed-book, open-domain generative question answering datasets.
  arXiv  Detail & Related papers  (2024-04-04T11:32:03Z)
• The Implicit Delta Method [61.36121543728134]
  In this paper, we propose an alternative, the implicit delta method, which works by infinitesimally regularizing the training loss of uncertainty. We show that the change in the evaluation due to regularization is consistent for the variance of the evaluation estimator, even when the infinitesimal change is approximated by a finite difference.
  arXiv  Detail & Related papers  (2022-11-11T19:34:17Z)
• Logical Satisfiability of Counterfactuals for Faithful Explanations in NLI [60.142926537264714]
  We introduce the methodology of Faithfulness-through-Counterfactuals. It generates a counterfactual hypothesis based on the logical predicates expressed in the explanation. It then evaluates if the model's prediction on the counterfactual is consistent with that expressed logic.
  arXiv  Detail & Related papers  (2022-05-25T03:40:59Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.