Shortcut Learning

Ask about this topic

Overview

Shortcut learning studies how models come to rely on spurious correlations, such as backgrounds, metadata, speaker identity, or dataset structure, rather than evidence relevant to the task. Researchers diagnose these behaviors with controlled interventions, counterfactual and cross-domain tests, modality ablations, attribution methods, and probes of learned representations. Mitigation approaches include data rebalancing and targeted augmentation, causal or adversarial objectives, feature disentanglement, architectural constraints, and post-training or inference-time interventions. A central challenge is that high accuracy and even readable representations do not establish what drives a decision, while suppressing a shortcut can also remove useful signal or fail when the confounder changes.

Latest papers

CardsList

159 papers25 papers in the last four weeks, up 108% on the four weeks before. 0.1% of all new papers.

Aug 2, 2026

Harnessing Adversarial Distillation to Customise Debiased, Disease-Specific Pathology Foundation Models for Breast Cancer

Pathology foundation models (PFMs) provide strong tissue representations and have become central to digital pathology. However, deployment in disease-specific settings is limited by 1) the high computational cost of billion-parameter PFMs and 2) distribution mismatch and non-biological bias inherited from pan-cancer, multi-centre pre-training, including site-specific signatures and imbalanced disease prevalence. These factors can encourage shortcut learning and under-emphasise subtle morphology required for reliable modelling of a specific cancer type. We present SmartStu (a Smart Student), a framework to customise compact, breast-cancer-specific PFMs via distillation whilst mitigating confounding. SmartStu distils representations from multiple teacher PFMs into a lightweight student backbone. Crucially, we introduce adversarial distillation that leverages a dedicated noise model trained to predict nuisance, edge-dominated cues on the distillation set. Using this noise model as a counterexample, the adversarial objective encourages the student to recognise, yet suppress, features predictive of nuisance targets. We further incorporate multi-teacher ensemble distillation and an auxiliary self-supervised objective with artefact injection. We validate SmartStu on three external cohorts (Yale HER2, SLN-Breast, and BRACS) with multiple tiny backbones. SmartStu yields breast-cancer-specific PFMs that are over 30×30\times smaller than general PFMs whilst largely preserving, and sometimes improving, downstream performance measured by balanced accuracy (bAcc) and AUC. Code is available at https://github.com/zwchen03/advDistall.

Zhiwei Chen, Yang Hu, Yuxiang Xiao +7

School of Computer Science and Engineering, South China University of Technology, China · School of Computing and Mathematical Sciences, University of Leicester, UK · Leicester Cancer Research Centre, University of Leicester, UK +4

Computational PathologyPathology Foundation ModelsOncologyShortcut LearningKnowledge DistillationAdversarial Knowledge Distillation
AskPDF
Jul 28, 2026

A Distributional Robustness Margin For Pathology Foundation Models

Pathology foundation models encode non-biological variation introduced by tissue preparation, staining and scanning, enabling shortcut learning that undermines generalisation across institutions. The Robustness Index (RI} was proposed to assess whether local representation geometry is dominated by biological or non-biological variation. However, its construction suffers from structural limitations that make cross-model comparison unreliable and call for a more principled metric. We introduce the Cross-confounder Robustness Margin (CRoMa), which measures, for each sample, whether biologically matched samples differing in the confounder lie closer in representation space than confounder-matched samples differing in biology. By design, CRoMa recasts robustness as a cohort-wide distribution of sample-level margins. We evaluated frozen representations from 20 tile-level encoders across three benchmarks and 4 slide-level encoders on a fourth. Median CRoMa rankings were broadly consistent across cohorts, yet all encoders contained confounder-dominated subsets whose prevalence and severity varied substantially across models and cohorts. Higher CRoMa margins were associated with smaller shortcut-induced performance drops after supervised adaptation, indicating that CRoMa can be a valuable tool for assessing model robustness on downstream tasks.

Clément Grisi, Jeroen van der Laak, Geert Litjens

Department of Pathology, Radboud University Medical Center, Nijmegen, The Netherlands

Distribution Shift RobustnessNeural Network RobustnessPathology Foundation ModelsShortcut Learning
AskPDF
Jul 27, 2026

Analysis of the Shortcut Learning and Clever Hans Effect in CNN based ECG Image Classification

Deep learning models for ECG image classification may achieve high accuracy by exploiting non-physiological visual cues instead of ECG waveform morphology. Given the black-box nature of deep learning models, their promise of high predictive performance often remains insufficiently translated into clinical or real-world trust, interpretability, and actionable decision-making. In this study, we examine shortcut learning and Clever Hans effect in a publicly available ECG image dataset using convolutional neural networks. In process we have created six image-derived feature sets (FSs), FS1: raw full ECG images, FS2: cropped waveform-only images, FS3: waveform-masked metadata images, FS4: red-arrow artifact images for the myocardial infarction class, FS5: contrast-enhanced images for the abnormal heartbeat class and FS6: Gaussian-blurred images for the normal class. These controlled representations were used to test whether classification performance persists when waveform information is removed or when artificial class-specific artifacts are introduced. Shortcut retention score, prediction consistency and confidence divergence across Feature-Set Representations have been calculated to assess the transparency about the learning pattern. Along with factual results, average Integrated Gradients and occlusion sensitivity test results are presented to inspect whether model attribution focused on ECG-relevant waveform regions or on non-clinical artifacts. Performance changes across feature sets and attribution patterns were used to identify potential Clever Hans behavior. This study evaluates whether ECG image classifiers learn clinically meaningful morphology or shortcut cues introduced by report layout, metadata, contrast, blur, or artificial markers.

Abhay Kumar Pathak, Mrityunjay Chaubey, Manjari Gupta +1

Department of Computer Science, Institute of Science, Banaras Hindu University, Varanasi, India. · University of Petroleum and Energy Studies, Dehradun, India. · Department of Computer Science (IDI), Norwegian University of Science and Technology (NTNU), Gjøvik, Norway.

Electrocardiogram ClassificationConvolutional Neural NetworksExplainable Medical Image AnalysisShortcut Learning
AskPDF
Jul 27, 2026

What do Reward Models Memorize?

This paper studies what discriminatively trained reward models (RMs) memorize by measuring counterfactual memorization on two human preference datasets. We show that RMs 1) misallocate memorization to easy, high margin preference pairs, 2) memorize dataset-specific shortcuts (e.g., model identity, user sampling strategy), and 3) overgeneralize simple heuristic correlates of human preference (e.g., length, compliance) when confronted with unseen preference pairs. Overall, our findings indicate that discriminative training of RMs from human preference data results in biased RMs not yet capable of judging response quality in context-dependent scenarios.

Ivo Verhoeven, Pushkar Mishra, Ekaterina Shutova

ILLC, University of Amsterdam, The Netherlands · Google DeepMind, London, United Kingdom

Reward ModelingPairwise Preference EvaluationHuman Preference ModelingShortcut LearningPreference Modeling
AskPDF
Jul 26, 2026

PathScale-R1: Cross-scale Reasoning for Pathological Image Analysis

Pathological diagnosis is inherently multi-scale, requiring the integration of global tissue architecture at low magnification with cellular morphology at higher magnification. However, existing pathology benchmarks and vision-language models (VLMs) are still largely developed under single-scale settings, limiting their ability to learn clinically meaningful multi-magnification reasoning. Moreover, naively constructed visual question answering (VQA) tasks may be susceptible to text-only or superficial visual shortcuts, leading to unreliable assessments of visual understanding. To address these limitations, we introduce a benchmark and training framework for shortcut-resistant cross-scale pathology reasoning. We design an Adversarial Text-only Screening strategy for semantic reasoning questions and a Structure-controlled Distractor Sampling strategy for visual grounding questions, encouraging models to rely on cross-scale visual evidence. Based on this pipeline, we construct PathScale-VQA, a high-quality cross-scale pathology VQA benchmark with 10,373 multiple-choice questions grounded in 1,368 diagnostic paths across multiple magnification levels. Building on the semantic reasoning set, PathScale-R1 is optimized through Difficulty-driven Reasoning Distillation supervised fine-tuning followed by reinforcement learning with a Scale-aware Reasoning Structure reward, which encourages the use of evidence across magnifications. Extensive experiments demonstrate state-of-the-art performance of PathScale-R1 on cross-scale reasoning tasks and effective transfer to conventional single-scale pathology VQA. Our code is available at https://github.com/iMVR-PL/PathScale-R1.

Chi Phan, Tianyi Zhang, Yufeng Wu +7

Department of Electrical and Computer Engineering, National University of Singapore, Singapore 117417 · Department of Biomedical Engineering, National University of Singapore, Singapore 117417, Singapore · Department of Pathology, Fujian Medical University Cancer Hospital & Fujian Cancer Hospital, Fuzhou, China +1

Computational PathologyMedical VQAVLM ReasoningShortcut LearningPathology VQAMedical Image Benchmarks
AskPDF
Jul 25, 2026

Breaking the Synthetic-Real Domain Shortcut for Training-Free Generative Replay-based Class Incremental Learning

Class-incremental learning (CIL) requires models to continuously acquire new knowledge while avoiding catastrophic forgetting. While exemplar replay is effective, it raises concerns regarding privacy and storage. Thus, generative replay has emerged as a viable alternative, synthesizing old data using frozen pretrained text-to-image (T2I) models without any extra training. However, we observe that directly mixing synthetic old-class data with real new-class data during incremental training leads to significant performance degradation. This issue stems from a "domain shortcut", where models rely on domain-discriminative features instead of semantic class cues. To address this, we propose DREAM (D‾\underline{\mathbf{D}}omain-R‾\underline{\mathbf{R}}egularized E‾\underline{\mathbf{E}}xemplar-free A‾\underline{\mathbf{A}}lignment M‾\underline{\mathbf{M}}odel), which uses a training-free generator to synthesize old-class data and eliminates domain shortcut via subspace rectification and orthogonal projection, while reinforcing semantic alignment through real-anchored prototype regularization. Extensive experiments on 4 datasets demonstrate that DREAM outperforms existing exemplar-free CIL methods and achieves state-of-the-art performance. Our source code is available at https://github.com/Light-ZhangTao/DREAM.

Tao Zhang, Qixuan Fan, Yiyuan Liang +7

Huazhong University of Science and Technology, Wuhan, China · State Key Laboratory of Multispectral Information Intelligent Processing Technology, Wuhan, China · University of Science and Technology of China, Anhui, China +1

Class-Incremental LearningSynthetic-to-Real Domain AdaptationReplay-Based Continual LearningIncremental LearningShortcut Learning
AskPDF
Jul 24, 2026

Alleviating Regional Shortcuts for Few-Shot Class-Incremental Learning

Few-shot class-incremental learning (FSCIL) aims to incrementally learn novel classes with only a few samples while avoiding forgetting base classes. However, current methods show a tendency to misclassify novel-class samples into base classes, which we find to be caused by the excessive focus on base-class-discriminative regions on novel-class samples. In this work, we aim to explore the underlying mechanism for an interpretation and solution. We first provide a compositional view to analyze the transferred and reused spatial patterns on novel-class samples. Then, through extensive experiments and theoretical analysis, we identify both empirically and theoretically that a shortcut exists in the model's base-class training, which naturally forms the excessive focus on only the most discriminative regions (primitives), which we term as the regional shortcut. Finally, based on this interpretation, to address this problem, we propose a compositional-learning-based method to learn two primitive sets (a common set and a discriminative set), which alleviates the regional shortcut by constraining the model to learn and utilize the common primitive set for base- and novel-class recognition. Extensive experiments on standard FSCIL benchmarks demonstrate the effectiveness of our approach, yielding consistent improvements over existing state-of-the-art methods in both accuracy and interpretability.

Haichen Zhou, Yazhe Lyu, Yixiong Zou +2

School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074, China

Class-Incremental LearningIncremental LearningFew-Shot Image ClassificationShortcut LearningCompositional Representation Learning
AskPDF
Jul 24, 2026

Listen, Do Not Copy: Internalizing Audio-Grounded Scaffold Context for Robust Omni-Model Speech Understanding

Omni models transcribe clean, single-speaker speech well, but their accuracy drops sharply when speakers overlap and the scene is noisy, exactly where knowing who said what matters most. A natural fix is a short scene description. We show why this is risky: answer-bearing text lets the model copy instead of listen, so the score rises although nothing has been heard; a silent test exposes this shortcut at once. We call this failure mode perception bypass and address it with Audio-Grounded Scaffold Context (AGSC). AGSC links three steps: first, we build clues from audio to guide listening without giving the answer; second, answer-overlap and silence tests probe them for leakage and audio dependence; finally, those clues scaffold training but vanish at test time, yielding no-clue capability. Across three heterogeneous Omni models, training on AGSC lowers no-clue capped mean permutation word error rate (mpWER) on overlapping, noisy speech from 25%-71% to 9%-15%. For streaming control, we formulate a joint GDPO task in which the model learns when to use a clue and how to produce a speaker-attributed transcript from separately normalized format, gate, and transcript rewards. After internalization, AGSC adds almost no inference overhead.

Pengfei Zhang, Biao Tian, Tianxin Xie +3

AI Thrust, Information Hub, The Hong Kong University of Science and Technology (Guangzhou) · Token Foundry, Alibaba Group, China

Speaker-Attributed ASRContextual ASRStreaming ASRAutomatic Speech RecognitionShortcut LearningSpeaker Diarization
AskPDF
Jul 23, 2026

Probing Speaker Identity Sensitivity in Audio Deepfake Detectors

Audio deepfake detectors are trained to distinguish genuine speech from synthetic speech and often perform well on standard benchmarks. Yet the same detector that achieves less than 1% error on one dataset can see its error rate increase twentyfold when evaluated on a different dataset. We argue that one contributing factor is speaker-identity reliance: standard training corpora correlate speaker identity with the genuine/synthetic label, allowing detectors to partially rely on speaker-related cues rather than synthesis artifacts alone. We propose the Identity Sensitivity Score (ISS), a per-utterance diagnostic that quantifies how much a detector's output changes across different speaker identity contexts. ISS requires no ground-truth labels at inference time and can be computed from the detector score and a pool of reference speaker examples. Across two detectors and two datasets, incorrectly classified utterances have ISS scores 29 to 52 times higher than correctly classified utterances, and ISS alone predicts misclassification with area-under-curve (AUC) up to 0.954. To test whether ISS actually captures identity-sensitive behavior rather than serving only as a proxy for prediction confidence, we apply voice conversion to 500 utterances and measure the resulting detector-score shift. Utterances flagged as identity-sensitive by ISS respond 19 to 30 times more strongly to this manipulation than utterances flagged as stable. These results position ISS as a practical inference-time diagnostic for speaker-dependent failure analysis in audio deepfake detection.

Daniyal Kabir Dar, Arun Ross

Michigan State University, USA

Audio Deepfake DetectionShortcut Learning
AskPDF
Jul 23, 2026

Detectors Learn the Wrong Thing: Shortcut-Resistant Adversarial Training Against Physically Realizable Attacks

AI-enabled visual perception systems are increasingly deployed in intelligent transportation infrastructure and autonomous vehicle related applications. However, physically realizable adversarial appearances pose a significant reliability challenge for these safety-critical systems. Adversarial training is effective, but repeated co-occurrence between adversarial texture and positive person instances can cause detectors to treat the texture itself as evidence of object presence, forming a patch texture shortcut. The detector may then treat texture as evidence for the target, causing false detections on texture-only inputs and weakening cross attack generalisation. We propose InsCAT, an instance-level contrastive adversarial training framework that prevents detectors from using adversarial texture as an independent decision cue. SICA aligns adversarial person features with matched clean features and separates them from texture-only negatives, while ROPO and Guard maintain online attack pressure and coordinate training. We evaluate eight independently generated attack textures on rendered nuScenes, INRIAPerson, printed garments, and three detector families. InsCAT achieves an average attack AP of 82.3% on rendered nuScenes, exceeding the strongest baseline by 11.1 points.Relative to AT-Mix, texture FPR decreases from 46.9% to 7.3%. Physical tests yield an F1 score of 96.6% and an FPR of 1.8%. Consistent gains across separately trained detectors demonstrate applicability across architectures with direct inference. The findings show that robust physical detection depends on preserving target related evidence while preventing adversarial texture from becoming an independent decision cu

Yuanhao Huang, Yilong Ren, Jinlei Wang +3

School of Transportation Science and Engineering, Beihang University, 100191, Beijing, P.R China · Zhongguancun Laboratory, 100191, Beijing, P.R China · School of Systems Science and Engineering, Sun Yat-Sen University, 510006, Guangzhou, P.R China +2

Adversarial TrainingAdversarial RobustnessObject DetectionShortcut LearningPhysical Adversarial AttacksAdversarial Attacks on Object Detectors
AskPDF
Jul 20, 2026

Style over Substance: A Shortcut Audit of Emotion-Description Preference Evaluation

Preference over model-generated emotion descriptions is emerging as a standard evaluation metric for multimodal emotion understanding, exemplified by the MER2026 MER-Prefer track on EmoPrefer. Such benchmarks assume that predicting the preferred description requires grounded cross-modal understanding of the video. We conduct a systematic shortcut audit of EmoPrefer using content-blind probes. A simple logistic regression using only description length and generator identity, without processing the text, video, or audio, performs comparably to LoRA-finetuned 7B text and audio-visual judges (65.8 versus 66.8 WAF on EmoPrefer-V2). Generator identity is recoverable from description text with 99.5 percent accuracy, every candidate pair contrasts two distinct generators, and the human preference labels agree with a fold-exclusive per-generator win-rate prior on 66 percent of the evaluated pairs. When the human label conflicts with this prior, trained judges still follow the style prior on 63 to 80 percent of the pairs. On a length-matched subset that neutralizes verbosity bias, the tested media configurations yield no statistically significant improvement, while an ODIN-inspired diagnostic that decouples the style shortcut leaves its content head near chance. These results do not imply that human preferences are inherently stylistic or that the descriptions contain no emotional information. Instead, they show that the current scores can be reached without verifying either description against the video. We recommend source-balanced pairing, strict length control, counter-stereotypical sliced reporting, and multi-annotator consensus for future cross-generator evaluations. Code is available at https://github.com/jiabingyang01/EmoPrefer-Audit.

Jiabing Yang, Yixiang Chen, Yuan Xu +6

University of Chinese Academy of Sciences Institute of Automation, Chinese Academy of Sciences Beijing, China

Human Preference EvaluationPairwise Preference EvaluationMultimodal Emotion RecognitionShortcut LearningMultimodal Model Evaluation
AskPDF
Jul 17, 2026

Controlling Implicit Shortcut Reliance in L2 Spoken English Auto-markers

Increasingly, speech and language processing tasks take either audio or text directly rather than extracting features from these as the input to the classifier or regressor. Often these systems make use of complex, for example transformer-based, processes that have the ability to derive highly non-linear mappings between the input and the output. Unfortunately these systems can also learn ''shortcuts'' where the classifier is overly reliant on particular aspects of the input to yield the output. For the task of language proficiency assessment, this over-reliance can enable learners to increase their score by exploiting the shortcut rather than improving their ability. This paper introduces a novel training criterion that is able to reduce the classifier's reliance on shortcuts, thus for example limiting this option for malpractice in language assessment. This process is illustrated on two forms of assessment system, one based on the audio the other on the speech recognition text. The results show that, for both systems, there is higher correlations with features that could be exploited for malpractice than expected from the human reference, indicating an over-reliance on these features. By introducing the modified training criterion, this correlation can be reduced to be closer to the reference correlation.

Shilin Gao, Mark J. F. Gales, Kate M. Knill

ALTA Institute, Machine Intelligence Laboratory, Cambridge University Engineering Department Cambridge, UK

Educational AssessmentAutomated EvaluationShortcut LearningAutomated Spoken Language Assessment
AskPDF
Jul 15, 2026

Genre Bias or Aesthetic Perception? Identifying and Mitigating Shortcut Learning in Music Evaluation

Music aesthetics scoring plays a critical role in applications such as dataset curation, generative model evaluation, and reward modeling for music generation. Recent approaches rely on deep neural networks trained on human-annotated ratings, but these models may exploit spurious correlations rather than capturing perceptually meaningful aesthetics. In this work, we identify a previously underexplored failure mode in music evaluation models: genre-induced shortcut learning. Through a systematic analysis of SongEval, we show that biases in training data lead to strong correlations between genre-related features and predicted scores, causing the model to use them as a proxy for aesthetics. This results in systematic overestimation of pop music and undervaluation of high-quality samples from other genres, leading to predictions that are inconsistent with human preferences. To address this issue, we propose a training objective that jointly reweights hard samples and regularizes group-level performance, encouraging the model to learn genre-invariant representations of musicality. Experimental results demonstrate that our method reduces genre-dependent bias and improves alignment with human preferences, as reflected by gains in both cross-genre and within-genre preference alignment.

Yizhou Zhang, Wangjin Zhou, Yi Zhao +3

Graduate School of Informatics, Kyoto University, Japan · WXG, Tencent, China

Human Preference EvaluationAlgorithmic BiasShortcut Learning
AskPDF
Jul 15, 2026

Auditing Protocol-Level Shortcuts in Large Audio Language Model Judges for Speech Evaluation

Large audio-language models (LALMs) are increasingly used as automatic judges for speech evaluation. However, high agreement with human ratings does not guarantee that their verdicts are grounded in the audio. A judge may instead rely on specialist labels or reference data supplied by the evaluation protocol itself, taking a shortcut in place of listening to the audio. In this paper, we audit such protocol-level ``shortcuts'' in LALM judges across three common deployment protocols: feature-blueprint judging, where the audio is replaced by a structured text description of acoustic features, reference-conditioned judging, and pairwise A/B comparison. Across six judges and four attributes, we find that several LALMs rely on protocol-level shortcuts. For example, in feature-blueprint judging, incorrect specialist labels reduce five judges' emotion accuracy to 0.10 or below, and in concatenated A/B comparisons, Qwen3-Omni-Thinking often picks the same slot regardless of order swaps. These results indicate that aggregate agreement can overstate the validity of LALM judges unless the model and the evaluation protocol are assessed jointly, and that each model-protocol pair should be evaluated with a matched shortcut probe.

Joonyong Park, David M. Chan, Yuki Saito +1

Graduate School of Information Science and Technology, The University of Tokyo, Tokyo, Japan · Berkeley AI Research (BAIR), University of California, Berkeley, CA, USA

Audio-Language Model EvaluationLLM AuditingSpeech Quality AssessmentAudio-Language ModelsShortcut Learning
AskPDF
Jul 14, 2026

Inhibited Self-Attention: Sharpening Focus in Vision Transformers

Vision Transformers (ViTs) have demonstrated remarkable performance in computer vision tasks. However, their self-attention mechanism often diffuses focus across background regions, relying on spurious correlations rather than object-relevant cues. Inspired by inhibitory mechanisms observed in biological vision systems, we propose the Inhibited Self-Attention (ISA), a novel self-attention that integrates inhibitory signals to enhance feature selectivity and suppress spurious responses. In contrast to conventional self-attention, which relies solely on positive attention values due to softmax normalization, our approach retains and utilizes negative attention scores to suppress irrelevant features and sharpen focus on objects of interest. Experiments across multiple datasets, including ImageNet-1k and COCO, and several robustness benchmarks demonstrate that ISA enhances object-centric selectivity, reduces shortcut reliance, and improves out-of-distribution generalization. Our analysis of relevance maps confirms that ViTs with ISA exhibit sharper, more localized focus on object-relevant regions while reducing distractions from non-relevant (background) features, enabling more reliable models. We release our code at https://github.com/prdvanderwal/inhibited-self-attention

Peter R. D. van der Wal, Nicola Strisciuglio, George Azzopardi

University of Groningen · University of Twente · 3Stellenbosch University +1

Self-AttentionVision TransformerShortcut Learning
AskPDF
Jul 13, 2026

Do Video-LLMs Actually Watch? Diagnosing Character-Tracking Failures in Long-Form Video

Can a Video Large Language Model (Video-LLM) follow one person through a long video, keeping track of who they are well enough to report, in order, how their outfit changes across a full TV episode? Benchmarks increasingly score this kind of task, and the strongest open-source 7--8B models now reach 37--38% on InfiniBench's global appearance task, which asks exactly that. But does that score come from tracking the named character, or from something easier? We test this with a nine-condition diagnostic protocol applied to three architecturally distinct open-source Video-LLMs, with Gemini2.5Flash as a frontier reference, and find the accuracy does not come from character tracking. When we change the character named in the question to a different cast member, leaving the video and answer options untouched, the models change their answer only 4--31% of the time, so they are largely ignoring who the question asks about. Breaking that test down by the gender of the swapped name shows why: the models react more when the name is changed to a different-gender character than to a same-gender one (a 13--28 point gap), picking up coarse gender cues but unable to tell same-gender individuals apart. This shallow processing surfaces again when we drop the multiple-choice options and ask the same questions open-endedly: open-source accuracy drops 18--25 points, with none of 151 answers fully correct, versus a 12-point drop for Gemini. Further checks rule out the obvious innocent explanations, adding subtitles, using the most informative frames, or doubling the number of frames all leave character tracking unimproved, so the bottleneck is not how much video the model sees but how it ties that video to the person the question names. We release a diagnostic toolkit for auditing what such benchmark scores actually measure.

Mohammad Al-Ratrout, Shayla Sharmin, Aditya Raikwar +1

Visual Object TrackingVideo-Language Model EvaluationVideo QALong-Video UnderstandingShortcut LearningMultimodal Model Evaluation+1
AskPDF
Jul 12, 2026

RED-Sphere: Hyperspherical Residual Edge Debiasing for Cross-Population Fundus Disease Domain Generalization

Medical image classifiers are often trained within one source population, yet clinical deployment requires robustness to patients whose appearance, acquisition style, and disease prevalence differ from the source cohort. Existing fairness and robustness methods often require group supervision or treat appearance variation as an undifferentiated nuisance, which is insufficient when population-correlated low-level cues and lesion evidence share edge and texture structure. We study a strict source-only cross-population setting, where external populations are unseen during optimization, validation, scheduling, hyperparameter and model selection. We propose RED-Sphere, a plug-and-play robustness framework for image classification under unseen population shifts. It estimates shortcut-sensitive nuisance responses with an edge and feature energy prior, attenuates dominant responses through residual soft gating, regularizes masked nuisance views with counterfactual-inspired consistency and separation losses, and predicts labels with normalized spherical prototypes. It favours angular semantic evidence over source-correlated activation magnitude while preserving lesion structure. Although demonstrated on 2D Scanning Laser Ophthalmoscopy (SLO) fundus classification for Age-Related Macular Degeneration (AMD) and Diabetic Retinopathy (DR), RED-Sphere is not tied to retinal anatomy: the same principle can be adapted with modality-specific nuisance priors wherever appearance shortcuts and semantic evidence are entangled. Under a strict White-only Harvard-FairVision protocol, RED-Sphere improves held-out macro-F1 across all 20 task and backbone comparisons, with average gains of 1.28 and 2.98 F1 points on AMD and DR. Gains in AUC and PR-AUC, visual diagnostics, ablations, and sensitivity analyses further support stronger external semantic alignment and more stable angular disease geometry.

Yan Lin, Ziheng Wang, Shuang Chen +2

Newcastle University Newcastle upon Tyne, UK · King Abdullah University of Science and Technology (KAUST) Thuwal, Saudi Arabia · Durham University Durham, UK

Distribution Shift RobustnessMedical Image ClassificationFundus ImagingShortcut LearningCross-Dataset Generalization
AskPDF
Jul 12, 2026

Artificial Foveated Perception for Mitigating Shortcut Learning in Robotic Foundation Models

Robotic foundation models still need task-specific fine-tuning before deployment, and the fine-tuned policies often break under modest changes in scene layout, lighting, or nearby distractors. We trace this brittleness to shortcut learning: fine-tuning supervises actions but not the visual evidence the policy uses, so the policy can settle on scene-level correlations that predict the demonstrations without causing success. We propose Artificial Foveated Perception (AFP), a lightweight, policy-agnostic module that takes the same vision and language inputs as existing Vision-Language-Action and World Action Model pipelines and predicts task-conditioned masks over the relevant objects, the robot, and other action-critical regions. During fine-tuning the masks serve as an auxiliary grounding signal that aligns the policy's visual attention with task-relevant regions; the policy architecture is unchanged, and at inference the policy runs on the original observation stream with no AFP call in the control loop. In simulation with four robotic foundation models and on a real robot with π0.5π_{0.5}, AFP improves generalization under environmental perturbations, reduces overfitting, and shortens fine-tuning. Ablations over mask quality and grounding-loss design show that these gains come from directing policy learning toward task-relevant visual evidence. Code, data, and videos are available at https://apollo-lab-yale.github.io/26-CoRL-AFP-website/.

Xiatao Sun, Yuan Zhuang, Mateo Sanchez Lopez Negrete +9

Yale University · University of Connecticut · Peking University +2

Robot Foundation ModelsVisual GroundingShortcut LearningRobustness of VLA ModelsVisual Grounding for RoboticsFoveated Vision
AskPDF
Jul 11, 2026

When Data Imbalance Helps: Robust Generalization Through Shortcut Saturation

We study robust generalization under spurious correlations: tasks where a shortcut feature is correlated with the true label in training but anti-correlated in an adversarial held-out split. Varying the spurious ratio rr (the fraction of training examples where shortcut = true label) and model capacity, we find a counterintuitive result: data imbalance promotes generalization in sufficiently capable models. On a synthetic task where the true label is sum parity of an integer sequence and the shortcut is the parity of the maximum-valued element, a 2-layer, 2-head transformer generalized (reached 100%100\% adversarial accuracy) in 0% of seeds at r=0.50r{=}0.50 but 77% of seeds at r=0.90r{=}0.90. The effect is absent in 1-layer models, where imbalance instead traps the model on the shortcut. Through mechanistic analysis -- gradient conflict dynamics, circuit evolution, and QK/OV circuit ablations -- we characterize a mechanistic pathway consistent with imbalance promoting generalization.

Cheng-Ting Chou, Duc Binh Hoang

University of California, Los Angeles

Class-Imbalanced LearningTransformerNeural Network RobustnessClassificationSpurious Correlation RobustnessShortcut Learning+3
AskPDF
Jul 10, 2026

Shortcut Trajectory Planning for Efficient Offline Reinforcement Learning

Diffusion-based trajectory planners have shown strong performance in offline reinforcement learning, but their iterative denoising process often incurs high inference cost. Consistency-based planners reduce the number of sampling steps, yet they typically rely on a two-stage teacher--student distillation pipeline that increases training cost and may introduce instability. We propose Shortcut Trajectory Planning (STP), an offline model-based reinforcement learning framework that incorporates shortcut models as efficient trajectory generators. STP trains a conditional shortcut trajectory model in a single stage, supports adjustable one-step and few-step inference through step-size conditioning, and selects candidate plans using a critic augmented with feasibility-aware correction. Across standard D4RL benchmarks, including locomotion, navigation, manipulation, and dexterous control tasks, STP achieves strong performance while simplifying the training pipeline for fast generative planning.

Guanquan Wang, Yoshimasa Tsuruoka

Department of Information and Communication Engineering · The University of Tokyo

Trajectory GenerationOffline RLMotion PlanningShortcut LearningModel-Based RL
AskPDF
Jul 10, 2026

Repairing Shape-Prior Shortcuts in Long-Range Single-Shot Fringe Projection Profilometry

Single-shot fringe projection profilometry (FPP) networks that regress depth directly can exploit a shape-prior shortcut, recovering depth from object boundaries rather than from fringe phase. On a photorealistic synthetic benchmark (15,600 fringe images, 50 objects at 1.5-2.1 m standoff), the best such UNet baseline plateaus at 14.54 mm object mean absolute error (MAE), and neither more data nor more capacity removes the shortcut, because neither changes the hypothesis space the optimizer searches. We introduce PhiCalNet, which outputs a wrapped-phase representation (sin⁡φ,cos⁡φ)(\sinφ, \cosφ) and maps it to depth through a fixed differentiable calibration layer, removing the shape-prior solution architecturally rather than by a loss penalty. Because the single-shot mapping is non-injective without fringe order, PhiCalNet takes the fringe order as auxiliary input, an assumption a sensitivity analysis shows tolerates realistic decoding error; a physics-informed (PINN) baseline with the same physics as a soft penalty yields no gain, isolating the architectural choice as the operative factor. PhiCalNet reduces object MAE 3.3x to 4.46 mm, its residual confined to 0.103% of pixels at the ±π\pmπ wrap discontinuity, and a three-frame extension reaches 1.16 mm. Two checks agree: interpretability makes phase the most decodable internal feature, and pixel-wise conformal uncertainty quantification, to our knowledge the first for FPP, localizes error at the same discontinuity, where rejecting the top 5% of pixels by snapshot disagreement cuts root-mean-square error by 64% versus 3.5% for the baseline.

Adam Haroon, Cody Fleming, Beiwen Li

Department of Mechanical Engineering, Iowa State University, 2529 Union Drive, Ames, 50011, Iowa, USA · College of Engineering, University of Georgia, 597 D. W. Brooks Drive, Athens, 30602, Georgia, USA

Fringe Projection ProfilometryMetric Depth Estimation3D ReconstructionShortcut Learning
AskPDF
Jul 8, 2026

ReMoDEx: A Local-to-Global Relevance-Based Model Decision Explainability Framework for large-Scale Image Datasets

Deep learning image classifiers achieve strong predictive performance yet remain opaque in how decisions are formed. A model may predict correctly while relying on irrelevant cues, shortcut associations, peripheral structures, or device level artifacts instead of task relevant regions. On large scale datasets this opacity is especially problematic, since inspecting heatmaps one sample at a time cannot scale to thousands of predictions. We propose Relevance Based Model Decision Explainability (ReMoDEx), a framework for systematic, dataset scale assessment of model decision behaviour in image classification. ReMoDEx defines a stepwise pipeline: model inference, target class selection, relevance map generation, heatmap standardisation, similarity based grouping of patterns, cluster level interpretation, and spatial relevance assessment. Local methods GradCAM++, Integrated Gradients, Occlusion Sensitivity, and Layerwise Relevance Propagation are each combined independently with a single global module that summarises an entire set of relevance maps into a few decision strategy clusters, replacing sample by sample inspection with an automatic, scalable summary. To demonstrate ReMoDEx, we applied it to a VGG16 based classifier distinguishing COVID-19, Normal, Lung Opacity, and Viral Pneumonia. The classifier showed stable performance (86.27% test accuracy, 0.9624 test AUC). However, each explainer combined with the global module consistently produced two recurring strategies: central thoracic region decisions and border/corner sensitive decisions, indicating possible shortcut learning that conventional metrics could not reveal. Masked image validation confirmed that model confidence and predicted class changed when central or peripheral regions were occluded. ReMoDEx thus provides a scalable relevance based decision assessment framework and an essential complement to accuracy based evaluation.

Abhay Kumar Pathak, Mrityunjay Chaubey, Manjari Gupta

Department of Computer Science, Institute of Science, Banaras Hindu University, Varanasi, India · University of Petroleum and Energy Studies, Dehradun, India

Explainable Artificial IntelligenceExplainable Image ClassificationExplainable Medical Image AnalysisShortcut LearningExplainability Evaluation
AskPDF
Jul 7, 2026

Association Restoration Test: Revealing Restorable Shortcuts after Unlearning

Association unlearning aims to disable learned label-attribute shortcuts while preserving task performance. Existing evaluations mainly measure output-level robustness or probe whether shortcut attributes remain readable in frozen features, but neither test determines whether a retained association remains functionally usable by the original classifier. We propose the Association Restoration Test (ART), a post-hoc diagnostic for functional shortcut restorability. ART estimates class-conditional association directions, amplifies residual components, and evaluates the modified features with the original classifier head. Across Waterbirds, CelebA, SpuCoDogs, and an ISIC timestamp-artifact extension, we show that output metrics, representation probes, and ART characterize distinct aspects of shortcut mitigation. These findings motivate restoration-aware evaluation for unlearning and shortcut-mitigation methods that target learned associations rather than individual classes or concepts.

Amy Lu, Changxiu Ji

Department of Computer Science, Stanford University

Spurious Correlation RobustnessShortcut LearningMachine UnlearningMachine Unlearning Evaluation
AskPDF
Jul 5, 2026

Shortcut Learning in Legal Judgment Prediction: Empirical Evidence from the UK Employment Tribunal

Current Legal Judgment Prediction (LJP) is constrained by its reliance on post-hoc judicial materials, increasing the likelihood that models perform retrospective classification rather than true forecasting. This paper empirically investigates shortcut learning in this context by studying claim-level outcome prediction in UK Employment Tribunal (UKET) decisions. Using a corpus of 33,158 individual claims, we predict outcomes from claim texts and LLM-extracted case summaries, evaluating models ranging from interpretable TF-IDF-based classifiers to black-box LLMs. While headline predictive performance figures appear strong, we demonstrate that such performance in LJP systems trained on post-hoc judicial text can be driven by the retrospective nature of the source material. Stratifying the test data by human judgments of leakage reveals that performance increases where outcome-revealing cues are embedded in the narrative. Moreover, a model trained on just the 4% of features identified as leakage achieves high performance, outperforming human experts. These findings substantiate concerns that LJP performance may be exaggerated by linguistic artefacts. Yet this vulnerability is not fatal to the research agenda. Instead, post-hoc judgments might be treated as potentially contaminated texts, requiring active auditing. Retraining models after masking leakage features results in only a negligible reduction in Macro-F1. Hence, while models will opportunistically exploit shortcuts when available, they remain capable of extracting useful predictive signals when these artefacts are removed.

Joe Watson, Joana Ribeiro de Faria, Marcus Tomalin +6

Faculty of Law, University of Cambridge, Cambridge, United Kingdom · The Psychometrics Centre, Cambridge Judge Business School, University of Cambridge, Cambridge, United Kingdom · Faculty of English, University of Cambridge, Cambridge, United Kingdom +3

Data LeakageLegal NLPLegal Judgment PredictionShortcut Learning
AskPDF
Jul 3, 2026

An Intervention-Based Framework for Shortcut Diagnosis in Spoofing Countermeasures

While deepfake audio detection systems achieve high performance in controlled benchmarks, their reliability often diminishes in the wild. Prior work shows that dataset-specific artifacts contribute to this gap. Yet, systematic tools to identify which acoustic properties a model exploits as shortcuts remain limited. We propose an intervention-based diagnostic framework, grounded in a directed graphical model, that formally distinguishes confound-driven shortcut dependencies from legitimate domain shift. We operationalise this through controlled acoustic perturbations targeting non-speech structure, spectral content, and signal energy, complemented by corpus-level distributional analysis. Evaluating XLS-R-300M with RawGAT-ST across ASVspoof challenges datasets, we quantify model sensitivity to specific intervention types. Results reveal that non-speech interventions produce the largest performance shifts, confirming non-speech intervals as a dominant shortcut.

Santiago Rubio, Pilar Bello, Dayana Ribas +3

ViVoLab, Arag´on Institute for Engineering Research (I3A), University of Zaragoza, Spain · BTS, Business Telecommunications Services, Spain

Audio Deepfake DetectionASVspoofShortcut LearningCausal Intervention
AskPDF
Jul 2, 2026

The Moving Eye: Enhancing VLA Spatial Generalization via Hybrid Dynamic Data Collection

Vision-Language-Action (VLA) models have shown remarkable promise in generalized robotic manipulation. However, their spatial generalization remains fragile. We argue that simply increasing the number of viewpoints is insufficient. Models often fall into the trap of Shortcut Learning, latching onto spurious correlations (e.g., fixed relative poses between objects or between the camera and robot base) rather than learning true spatial relationships. In this work, we propose a data-centric solution to enhance VLA spatial generalization. We utilize a dual-arm setup where one arm performs manipulation while the other serves as a mobile environmental camera. We systematically evaluate three data distribution patterns: Fixed, Multi-Fixed, and Moving Views. Our findings reveal that a hybrid strategy, combining continuous camera motion with diverse static viewpoints, yields the best performance by substantially reducing spurious correlations while maintaining training stability. Our experiments demonstrate that this strategy mitigates spurious correlations, enabling VLAs to generalize to unseen camera poses and object configurations where simply adding more static viewpoints fails. Crucially, we reveal that the susceptibility to shortcut learning and the struggle with spatial generalization are universal characteristics shared across diverse architectures. Consequently, all evaluated models (ACT, Diffusion, and VLA models including Pi0 and Gr00t) benefit significantly from our mixed data strategy.

Jincheng Tang, Yilong Zhu, Zhengyuan Xie +2

Lion Rock AI Lab, China Merchants Research Institute of Advanced Technology. · The Hong Kong University of Science and Technology. · Nankai University.

Generalization in Robotic ManipulationRobotic ManipulationVision-Language-Action ModelsRobot Policy GeneralizationShortcut LearningVision-Based Robot Control
AskPDF
Jul 1, 2026

Dataset Biases and Shortcut Learning in Motion-Based AI-Generated Video Detection

The visual quality of AI-generated videos has improved drastically in recent years, making it increasingly difficult for humans to distinguish between real and synthetic media. In this work, we evaluate the robustness and applicability of four state-of-the-art motion-based AI-generated video detectors. We identify significant preprocessing and sampling biases in three of the four methods and demonstrate that they account for a substantial portion of their reported performance. Furthermore, we find that these detectors are highly sensitive to motion patterns specific to their evaluation datasets, where AI-generated videos generally exhibit less inter-frame movement than real videos. We show that for all detectors, performance collapses to near-random levels when evaluated on a dataset that does not contain this motion bias. Additionally, through dataset rebalancing and the application of simple spatial augmentations, we observe severe performance degradation across all evaluated models. In contrast, we find that an existing frequency-based detector maintains strong performance across all evaluated datasets, suggesting that frequency-based approaches may offer a more generalizable path forward for AI-generated video detection. We hope that our work raises awareness towards these vulnerabilities and encourages the development of more representative, unbiased datasets and more robust evaluation protocols.

Joren Michels, Lode Jorissen, Nick Michiels

Digital Future Lab, Flanders Make, Hasselt University

Algorithmic BiasDeepfake DetectionShortcut LearningDataset Bias
AskPDF
Jul 1, 2026

Cross-Domain Generalization Failure in Lightweight Intrusion Detection Models for IIoT Networks

Lightweight machine learning models are increasingly proposed for intrusion detection in Industrial Internet of Things (IIoT) networks due to their suitability for resource-constrained edge deployment. Most reported results evaluate these models only within their training network, leaving behavior on unseen networks unverified. This study trains four lightweight architectures on one IIoT dataset and evaluates them, without retraining, on two structurally distinct IIoT datasets using a feature representation restricted to attributes available across all three sources. Explainability analysis across two top-performing models shows both rely overwhelmingly on coarse port-category features; the most influential category occurs in source-domain attack traffic at 96 to 435 times the rate in the two target domains, indicating that coarsening port resolution relocates rather than removes a documented shortcut. Evaluation under naturally imbalanced class distributions reveals a further effect: the evaluation protocol used can reverse which target network appears to pose the greater generalization challenge. Adversarial robustness and recovery through limited target-domain exposure are also assessed; robustness to adversarial perturbation is unrelated to cross-network generalization, and recovery through adaptation varies considerably by architecture. These findings suggest deployment readiness should be assessed using cross-network evaluation under realistic class distributions, rather than within-domain accuracy alone.

MD Azizul Hakim, Md Shihab Uddin, Talha Ibne Anis

Department of Computer Science and Technology, Bangladesh Sweden Polytechnic Institute Kaptai, Rangamati, Chittagong 4530, Bangladesh · Department of Electrical Engineering, Bangladesh Sweden Polytechnic Institute Kaptai, Rangamati, Chittagong 4530, Bangladesh

IoT Intrusion DetectionNetwork Intrusion DetectionCross-Domain GeneralizationEdge MLShortcut LearningIndustrial Internet of Things
AskPDF
Jun 29, 2026

SGD Provably Prioritizes a Shortcut Spurious Feature in the XOR Model

Neural networks are known to be susceptible to over-reliance on spurious correlations. However, the precise mechanism by which models exploit shortcut features is not fully understood, and algorithms to mitigate this behavior rely on as yet unjustified assumptions about the learned representations. In this work, we provide the first end-to-end theoretical characterization of spurious feature learning for two-layer ReLU neural networks trained by online minibatch SGD on the logistic loss. We consider data drawn from the high-dimensional Boolean hypercube with a quadratic signal function (namely XOR) and a linear spurious correlation. We show that SGD learns the spurious feature first, and exponentially fast. Moreover, the optimization dynamics couple the spurious and signal features, with a stronger spurious component inhibiting signal feature learning. Our analysis reveals precise phase transitions in the learning dynamics. In the first phase, alignment between the signs of the spurious feature and second-layer weight drives rapid growth of the spurious feature. In the second phase, large majority group margin slows learning and the signal feature remains suppressed. When the spurious correlation is maximally strong, we show theoretically that the spurious feature dominates even at the sample complexity threshold where XOR would be learned in isolation (i.e., if the spurious feature was absent). In contrast, when the correlation strength is constant, we provide preliminary empirical evidence that the model can eventually learn the XOR signal, although the spurious feature is not forgotten.

Tyler LaBonte, Vidya Muthukumar

H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology · School of Electrical and Computer Engineering, Georgia Institute of Technology

Representation LearningReLU Neural NetworksNeural Network Training DynamicsShortcut LearningStochastic Gradient DescentSpurious Correlation
AskPDF
Jun 29, 2026

CouCE: A Unified Causal Framework for Debiased Deep Metric Learning

Deep Metric Learning (DML) often struggles with zero-shot generalization because standard objectives inherently capture what co-occurs rather than what causes similarity. Consequently, DML models are vulnerable to shortcut learning driven by two structurally distinct confounders: background spurious correlations (which create backdoor paths via scene context) and foreground nuisance perturbations (which inject non-semantic variations like pose or illumination). Although existing methods have proposed targeted solutions for each pathway individually, none can simultaneously address both due to their fundamentally distinct causal roles. To bridge this gap, we propose the Counterfactual Causal Embedding (CouCE), a unified causal framework that explicitly models and neutralizes both confounders. Specifically, we introduce Orthogonal Dictionary-Based Backdoor Adjustment (ODBA), which isolates spurious background patterns into a variance-gated dictionary and stably disentangles them from the learned embeddings via soft orthogonal regularization. Simultaneously, we propose Multi-Scale Randomized Causal Intervention (MSRCI) to enforce causal invariance against foreground nuisances through multi-scale Fourier amplitude randomization and a symmetric KL invariance constraint. Notably, CouCE seamlessly integrates with any proxy-based loss, incurring modest training overhead without requiring architectural modifications during inference. Extensive experiments on CUB-200-2011, Cars-196, and Stanford Online Products demonstrate that CouCE consistently achieves state-of-the-art performance, providing a principled and robust solution for debiased DML.

Xin Yuan, Zhenyang Niu, Meiqi Wan +3

Wuhan University of Science and Technology Wuhan, China · Harbin Institute of Technology Harbin, China

Debiased MLCausal Representation LearningMetric LearningShortcut Learning
AskPDF