Decomposition of Evidence, Contradiction, and Fragility in Perturbation Responses

Perturbation methods explain model decisions by measuring prediction changes under altered inputs, but response magnitude tells us only how much a model reacts, not what that reaction means. The same magnitude can support the final factual-counterfactual difference, oppose it, or arise strongly along the perturbation path yet vanish at the endpoint. We therefore track how the contrast develops as paired inputs are progressively revealed, using the final contrast to interpret the trajectory. We introduce DECAF (Decomposition of Evidence, Contradiction, And Fragility), which routes aligned, opposed, and endpoint-null responses into evidence E, contradiction C, and fragility F. The decomposition preserves ordinary magnitude exactly, Abs = E + C + F, and is unique under endpoint-relative axioms. Across controlled vision and tabular settings, the three components track independently measured behavior. In a 72-model ImageNet-9 audit, we compare cases with nearly identical response magnitude but different independently measured behaviors. The largest DECAF component agrees with an observed behavior in 96.4% of cases, compared with 35.0% for magnitude alone. Changing only the reveal path increases total response by nearly 80%, yet evidence barely changes while fragility grows by more than 4x. On FunnyBirds and ImageNet-1k, short forward-only DECAF trajectories outperform the tested general-purpose attribution baselines. On a 1B-scale DINOv2 model, a short trajectory matches a strong gradient-based baseline with 4.75x lower wall time and 2.36x lower peak memory.

Best-of-Evidence: Best-of-N Selection under Partial Verification

BoN improves model outputs by sampling several candidates and selecting one with a proxy score, but it assumes that complete candidates can be evaluated reliably. Many vision-language tasks instead provide only partial verification: a finding, span, value, region, or relation may be checkable even when no dependable whole-response verifier exists. Moreover, the same claim may recur across candidates with opposing stances, allowing one observation to support part of the pool and contradict another. We introduce Best-of-Evidence (BoE), an inference-time selection framework that keeps the BoN candidate pool fixed, represents reusable claims with a signed candidate--factor graph, and allocates a limited budget to evidence actions that can change the final choice. BoE formalizes selection under partial verification and provides a practical score-based controller, with the zero-budget case recovering the underlying BoN decision. Theoretically, we show that residual evidence capacity limits any evidence-driven improvement and that shared factor queries can achieve an O(log K) versus Θ(K) query separation in a factor-code model. Common-ledger experiments on four medical VQA settings show that BoE can improve fixed-pool selection and rescue some BoN failures when evidence is reliable, contrastive, and decision-relevant, while also revealing the channel-quality and candidate-generation limits that prevent universal gains.

MedCore: Boundary-Preserving Medical Core Pruning for MedSAM

Medical segmentation foundation models such as SAM and MedSAM provide strong prompt-driven segmentation, but their image encoders are still too large for many clinical settings. Compression is also risky in medicine because a model can keep high Dice while losing boundary fidelity. We propose MedCore, a structured pruning framework for MedSAM. The main idea is to preserve two kinds of structures: structures that became important during SAM-to-MedSAM adaptation, and structures that have high boundary leverage. We identify the first type by a dual-intervention score that compares zeroing a group with resetting it to its original SAM weight. We identify the second type by boundary-aware Fisher estimation. We also introduce a boundary leverage principle, which shows that compression-induced boundary displacement is controlled by logit perturbation on the boundary divided by the logit spatial gradient. This principle explains why boundary metrics can degrade even when Dice remains high. On polyp segmentation benchmarks, MedCore reduces parameters by 60.0% and FLOPs by 58.4% while achieving Dice 0.9549, Boundary F1 0.6388, and HD95 5.14 after recovery fine-tuning. It also reaches 86.6% parameter reduction and 90.4G FLOPs with strong boundary quality. Our analysis further shows that MedSAM lies in a head-fragile boundary regime: head-pruning steps have 2.887 times larger 95th-percentile boundary leverage than MLP-pruning steps, and this logit-level effect is consistent with BF1 and HD95 degradation. Our code is available at https://github.com/cenweizhang/MedCore.

DISCOVER: A Solver for Distributional Counterfactual Explanations

Counterfactual explanations (CE) explain model decisions by identifying input modifications that lead to different predictions. Most existing methods operate at the instance level. Distributional Counterfactual Explanations (DCE) extend this setting by optimizing an optimal transport objective that balances proximity to a factual input distribution and alignment to a target output distribution, with statistical certification via chance constrained bounds. However, DCE relies on gradient based optimization, while many real-world tabular pipelines are dominated by non-differentiable models. We propose DISCOVER, a model-agnostic solver for distributional counterfactual explanations. DISCOVER preserves the original DCE objective and certification while replacing gradient descent with a sparse propose-and-select search paradigm. It exploits a sample-wise decomposition of the transport objective to compute per-row impact scores and enforce a top-$k$ intervention budget, focusing edits on the most influential samples. To guide candidate generation without predictor gradients, DISCOVER introduces an OT-guided cone sampling primitive driven by input-side transport geometry. Experiments on multiple tabular datasets demonstrate strong joint alignment of input and output distributions, extending distributional counterfactual reasoning to modern black box learning pipelines. A code repository is available at https://github.com/understanding-ml/DCE.

Axiomatic On-Manifold Shapley via Optimal Generative Flows

Shapley-based attribution is critical for post-hoc XAI but suffers from off-manifold artifacts due to heuristic baselines. While generative methods attempt to address this, they often introduce geometric inefficiency and discretization drift. We propose a formal theory of on-manifold Aumann-Shapley attributions driven by optimal generative flows. We prove a representation theorem establishing the gradient line integral as the unique functional satisfying efficiency and geometric axioms, notably reparameterization invariance. To resolve path ambiguity, we select the kinetic-energy-minimizing Wasserstein-2 geodesic transporting a prior to the data distribution. This yields a canonical attribution family that recovers classical Shapley for additive models and admits provable stability bounds against flow approximation errors. By reframing baseline selection as a variational problem, our method experimentally outperforms baselines, achieving strict manifold adherence via vanishing Flow Consistency Error and superior semantic alignment characterized by Structure-Aware Total Variation. Our code is on https://github.com/cenweizhang/OTFlowSHAP.

xai-cola: A Python library for sparsifying counterfactual explanations

Counterfactual explanation (CE) is an important domain within post-hoc explainability. However, the explanations generated by most CE generators are often highly redundant. This work introduces an open-source Python library xai-cola, which provides an end-to-end pipeline for sparsifying CEs produced by arbitrary generators, reducing superfluous feature changes while preserving their validity. It offers a documented API that takes as input raw tabular data in pandas DataFrame form, a preprocessing object (for standardization and encoding), and a trained scikit-learn or PyTorch model. On this basis, users can either employ the built-in or externally imported CE generators. The library also implements several sparsification policies and includes visualization routines for analysing and comparing sparsified counterfactuals. xai-cola is released under the MIT license and can be installed from PyPI. Empirical experiments indicate that xai-cola produces sparser counterfactuals across several CE generators, reducing the number of modified features by up to 50% in our setting. The source code is available at https://github.com/understanding-ml/COLA.

Epistemic Throughput: Fundamental Limits of Attention-Constrained Inference

Recent generative and tool-using AI systems can surface a large volume of candidates at low marginal cost, yet only a small fraction can be checked carefully. This creates a decoder-side bottleneck: downstream decision-makers must form reliable posteriors from many public records under scarce attention. We formalize this regime via Attention-Constrained Inference (ACI), in which a cheap screening stage processes $K$ records and an expensive verification stage can follow up on at most $B$ of them. Under Bayes log-loss, we study the maximum achievable reduction in posterior uncertainty per window, which we call \emph{epistemic throughput}. Our main result is a ``JaKoB'' scaling law showing that epistemic throughput has a baseline term that grows linearly with verification and prevalence, and an additional \emph{information-leverage} term that scales as $\sqrt{JKB}$, where $J$ summarizes screening quality. Thus, expanding cheap screening can nonlinearly amplify scarce verification, even when informative records are rare. We further show that this scaling is tight in a weak-screening limit, and that in the sparse-verification regime ($B \ll K$), substantial leverage requires heavy-tailed score distributions; for light-tailed scores the amplification is only logarithmic.

Quantifying Model Uniqueness in Heterogeneous AI Ecosystems

As AI systems evolve from isolated predictors into complex, heterogeneous ecosystems of foundation models and specialized adapters, distinguishing genuine behavioral novelty from functional redundancy becomes a critical governance challenge. Here, we introduce a statistical framework for auditing model uniqueness based on In-Silico Quasi-Experimental Design (ISQED). By enforcing matched interventions across models, we isolate intrinsic model identity and quantify uniqueness as the Peer-Inexpressible Residual (PIER), i.e. the component of a target's behavior strictly irreducible to any stochastic convex combination of its peers, with vanishing PIER characterizing when such a routing-based substitution becomes possible. We establish the theoretical foundations of ecosystem auditing through three key contributions. First, we prove a fundamental limitation of observational logs: uniqueness is mathematically non-identifiable without intervention control. Second, we derive a scaling law for active auditing, showing that our adaptive query protocol achieves minimax-optimal sample efficiency ($dσ^2γ^{-2}\log(Nd/δ)$). Third, we demonstrate that cooperative game-theoretic methods, such as Shapley values, fundamentally fail to detect redundancy. We implement this framework via the DISCO (Design-Integrated Synthetic Control) estimator and deploy it across diverse ecosystems, including computer vision models (ResNet/ConvNeXt/ViT), large language models (BERT/RoBERTa), and city-scale traffic forecasters. These results move trustworthy AI beyond explaining single models: they establish a principled, intervention-based science of auditing and governing heterogeneous model ecosystems.

Preventing the Collapse of Peer Review Requires Verification-First AI

This paper argues that AI-assisted peer review should be verification-first rather than review-mimicking. We propose truth-coupling, i.e. how tightly venue scores track latent scientific truth, as the right objective for review tools. We formalize two forces that drive a phase transition toward proxy-sovereign evaluation: verification pressure, when claims outpace verification capacity, and signal shrinkage, when real improvements become hard to separate from noise. In a minimal model that mixes occasional high-fidelity checks with frequent proxy judgment, we derive an explicit coupling law and an incentive-collapse condition under which rational effort shifts from truth-seeking to proxy optimization, even when current decisions still appear reliable. These results motivate actions for tool builders and program chairs: deploy AI as an adversarial auditor that generates auditable verification artifacts and expands effective verification bandwidth, rather than as a score predictor that amplifies claim inflation.

Fairness Research For Machine Learning Should Integrate Societal Considerations

Enhancing fairness in machine learning (ML) systems is increasingly important nowadays. While current research focuses on assistant tools for ML pipelines to promote fairness within them, we argue that: 1) The significance of properly defined fairness measures remains underestimated; and 2) Fairness research in ML should integrate societal considerations. The reasons include that detecting discrimination is critical due to the widespread deployment of ML systems and that human-AI feedback loops amplify biases, even when only small social and political biases persist.