Smaller self-supervised ViTs localize objects better via attention than larger ViTs, enabling A² to decouple localization from feature extraction for competitive performance on distribution-shifted benchmarks.
Last layer re-training is sufficient for robustness to spurious correlations
10 Pith papers cite this work, alongside 32 external citations. Polarity classification is still indexing.
10
Pith papers citing it
32
external citations · external index
citation-role summary
background 1
baseline 1
citation-polarity summary
representative citing papers
Proposes CBCM for diffusion-based spurious attribute mining and DCD for cross-projection debiasing, claiming SOTA worst-group accuracy on four benchmarks while tuning at most 0.22% of parameters.
Proposes memorization-guided two-stage scoring to select debiased training subsets, enabling ERM models to achieve better performance than SOTA debiasing techniques using only 10% of data.
Introduces MOOD benchmark for OOD LLM alignment failures and shows guard models plus Mahalanobis and perplexity OOD detectors improve recall from 39% to 45% with positive scaling.
DeconDTN-Toolkit simulates provenance shifts to expose ERM vulnerabilities and provides tools plus a robust OOD indicator for mitigating confounding by data provenance.
Characterizes spurious correlation mechanisms in preference optimization via mean spurious bias and causal-spurious correlation leakage, demonstrates irreducible vulnerability to distribution shift, and introduces tie training as selective mitigation with validation on log-linear models and empirica
Exploiting linear structure in VLM embeddings, a synthetic-data pre-training method yields background-invariant representations that exceed 90% worst-group accuracy on Waterbirds even under 100% spurious correlation with no minority examples in training.
Benchmark shows that combining data rebalancing with feature disentanglement mitigates shortcut learning more effectively than rebalancing alone in medical imaging models.
Sparse feature circuits are introduced as interpretable causal subnetworks in language models, supporting unsupervised discovery of thousands of circuits and a method called SHIFT to improve classifier generalization by ablating irrelevant features.
A multimodal fusion plus gradient-reversal unlearning framework improves MCI classification accuracy and reduces performance gaps across sex and language subgroups on TAUKADIAL and PREPARE.
citing papers explorer
-
$A^2$: Smaller Self-Supervised ViTs Localize Better than Larger Ones
Smaller self-supervised ViTs localize objects better via attention than larger ViTs, enabling A² to decouple localization from feature extraction for competitive performance on distribution-shifted benchmarks.
-
Dual-Branch Cross-Projection Debiasing through Diffusion-based Disentanglement
Proposes CBCM for diffusion-based spurious attribute mining and DCD for cross-projection debiasing, claiming SOTA worst-group accuracy on four benchmarks while tuning at most 0.22% of parameters.
-
Mitigating Spurious Correlations with Memorization-Guided Dataset De-Biasing
Proposes memorization-guided two-stage scoring to select debiased training subsets, enabling ERM models to achieve better performance than SOTA debiasing techniques using only 10% of data.
-
Benchmarking and Improving Monitors for Out-Of-Distribution Alignment Failure in LLMs
Introduces MOOD benchmark for OOD LLM alignment failures and shows guard models plus Mahalanobis and perplexity OOD detectors improve recall from 39% to 45% with positive scaling.
-
DeconDTN-Toolkit: A Library for Evaluation and Enhancement of Robustness to Provenance Shift
DeconDTN-Toolkit simulates provenance shifts to expose ERM vulnerabilities and provides tools plus a robust OOD indicator for mitigating confounding by data provenance.
-
Spurious Correlation Learning in Preference Optimization: Mechanisms, Consequences, and Mitigation via Tie Training
Characterizes spurious correlation mechanisms in preference optimization via mean spurious bias and causal-spurious correlation leakage, demonstrates irreducible vulnerability to distribution shift, and introduces tie training as selective mitigation with validation on log-linear models and empirica
-
Birds of a Feather Flock Together: Background-Invariant Representations via Linear Structure in VLMs
Exploiting linear structure in VLM embeddings, a synthetic-data pre-training method yields background-invariant representations that exceed 90% worst-group accuracy on Waterbirds even under 100% spurious correlation with no minority examples in training.
-
Mitigating Shortcut Learning via Feature Disentanglement in Medical Imaging: A Benchmark Study
Benchmark shows that combining data rebalancing with feature disentanglement mitigates shortcut learning more effectively than rebalancing alone in medical imaging models.
-
Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models
Sparse feature circuits are introduced as interpretable causal subnetworks in language models, supporting unsupervised discovery of thousands of circuits and a method called SHIFT to improve classifier generalization by ablating irrelevant features.
-
Fair Cognitive Impairment Detection Through Unlearning
A multimodal fusion plus gradient-reversal unlearning framework improves MCI classification accuracy and reduces performance gaps across sex and language subgroups on TAUKADIAL and PREPARE.