87 citations · 222 across the 32 of their papers we have counts for
38 papers
COFFEQA: Counterfactual-Guided Few-Shot Distillation for LLM Question Answering
Kavinda Kehelella, Sanghamitra Dutta
Knowledge distillation aims to train a smaller student model to mimic a larger teacher model for resource-constrained environments. However, the process of distillation is typicall…
Don't Judge an LLM Only by Its Activations: Discovering Suppressed Safety Features via Counterfactual Activation Potential
Swadesh Swain, Sanghamitra Dutta
Mechanistic interpretability has emerged as the primary means to understand safety behavior of LLMs. However, existing tools primarily focus on the activating neurons or features o…
Mechanistic Interpretability Reveals Shared Causal Subspaces in Brain-to-Speech Decoders
Maryam Maghsoudi, Ayushi Mishra, Sanghamitra Dutta
Decoding covert speech, such as mimed or imagined, from brain activity is harder than decoding vocalized speech. Cross-modal transfer, where information from one speech form helps…
Do Student LLMs Inherit OOD Robustness? Invariance-Weighted Distillation for Reliable Knowledge Transfer
Dileesha Kannangara, Sanghamitra Dutta
Knowledge distillation (KD) aims to compress high-performance teacher LLMs into lightweight students. However, distilled students often exhibit substantial performance degradation…
TabDistill: Distilling Transformers into Neural Nets for Few-Shot Tabular Classification
Pasan Dissanayake, Sanghamitra Dutta
Transformer-based models have shown promising performance on tabular data compared to their classical counterparts such as neural networks and Gradient Boosted Decision Trees (GBDT…
Few-Shot Knowledge Distillation of LLMs With Counterfactual Explanations
Faisal Hamman, Pasan Dissanayake, Yanjun Fu +1
Knowledge distillation is a promising approach to transfer capabilities from complex teacher models to smaller, resource-efficient student models that can be deployed easily, parti…
Improving Consistency in Retrieval-Augmented Systems with Group Similarity Rewards
Faisal Hamman, Chenyang Zhu, Anoop Kumar +4
RAG systems are increasingly deployed in high-stakes domains where users expect outputs to be consistent across semantically equivalent queries. However, existing systems often exh…
What If, But Privately: Private Counterfactual Retrieval
Shreya Meel, Mohamed Nomeir, Pasan Dissanayake +2
Transparency and explainability are two important aspects to be considered when employing black-box machine learning models in high-stake applications. Providing counterfactual exp…
T-SHIRT: Token-Selective Hierarchical Data Selection for Instruction Tuning
Yanjun Fu, Faisal Hamman, Sanghamitra Dutta
Instruction tuning is essential for Large Language Models (LLMs) to effectively follow user instructions. To improve training efficiency and reduce data redundancy, recent works us…
Counterfactual Explanations for Model Ensembles Using Entropic Risk Measures
Erfaun Noorani, Pasan Dissanayake, Faisal Hamman +1
Counterfactual explanations indicate the smallest change in input that can translate to a different outcome for a machine learning model. Counterfactuals have generated immense int…
Private Counterfactual Retrieval With Immutable Features
Shreya Meel, Pasan Dissanayake, Mohamed Nomeir +2
In a classification task, counterfactual explanations provide the minimum change needed for an input to be classified into a favorable class. We consider the problem of privately r…
Quantifying Knowledge Distillation Using Partial Information Decomposition
Pasan Dissanayake, Faisal Hamman, Barproda Halder +3
Knowledge distillation deploys complex machine learning models in resource-constrained environments by training a smaller student model to emulate internal representations of a com…
Private Counterfactual Retrieval
Mohamed Nomeir, Pasan Dissanayake, Shreya Meel +2
Transparency and explainability are two extremely important aspects to be considered when employing black-box machine learning models in high-stake applications. Providing counterf…
Quantifying Prediction Consistency Under Fine-Tuning Multiplicity in Tabular LLMs
Faisal Hamman, Pasan Dissanayake, Saumitra Mishra +2
Fine-tuning LLMs on tabular classification tasks can lead to the phenomenon of fine-tuning multiplicity where equally well-performing models make conflicting predictions on the sam…
Towards Formalizing Spuriousness of Biased Datasets Using Partial Information Decomposition
Barproda Halder, Faisal Hamman, Pasan Dissanayake +3
Spuriousness arises when there is an association between two or more variables in a dataset that are not causally related. In this work, we propose an explainability framework to p…
A Unified View of Group Fairness Tradeoffs Using Partial Information Decomposition
Faisal Hamman, Sanghamitra Dutta
This paper introduces a novel information-theoretic perspective on the relationship between prominent group fairness notions in machine learning, namely statistical parity, equaliz…
Model Reconstruction Using Counterfactual Explanations: A Perspective From Polytope Theory
Pasan Dissanayake, Sanghamitra Dutta
Counterfactual explanations provide ways of achieving a favorable model outcome with minimum input perturbation. However, counterfactual explanations can also be leveraged to recon…
REFRESH: Responsible and Efficient Feature Reselection Guided by SHAP Values
Shubham Sharma, Sanghamitra Dutta, Emanuele Albini +3
Feature selection is a crucial step in building machine learning models. This process is often achieved with accuracy as an objective, and can be cumbersome and computationally exp…
Demystifying Local and Global Fairness Trade-offs in Federated Learning Using Partial Information Decomposition
Faisal Hamman, Sanghamitra Dutta
This work presents an information-theoretic perspective to group fairness trade-offs in federated learning (FL) with respect to sensitive attributes, such as gender, race, etc. Exi…
Robust Counterfactual Explanations for Neural Networks With Probabilistic Guarantees
Faisal Hamman, Erfaun Noorani, Saumitra Mishra +2
There is an emerging interest in generating robust counterfactual explanations that would remain valid if the model is updated or changed even slightly. Towards finding robust coun…
Hyper-parameter Tuning for Fair Classification without Sensitive Attribute Access
Akshaj Kumar Veldanda, Ivan Brugere, Sanghamitra Dutta +2
Fair machine learning methods seek to train models that balance model performance across demographic subgroups defined over sensitive attributes like race and gender. Although sens…
Can Querying for Bias Leak Protected Attributes? Achieving Privacy With Smooth Sensitivity
Faisal Hamman, Jiahao Chen, Sanghamitra Dutta
Existing regulations prohibit model developers from accessing protected attributes (gender, race, etc.), often resulting in fairness assessments on populations without knowing thei…
Robust Counterfactual Explanations for Tree-Based Ensembles
Sanghamitra Dutta, Jason Long, Saumitra Mishra +2
Counterfactual explanations inform ways to achieve a desired outcome from a machine learning model. However, such explanations are not robust to certain real-world changes in the u…
Fairness via In-Processing in the Over-parameterized Regime: A Cautionary Tale
Akshaj Kumar Veldanda, Ivan Brugere, Jiahao Chen +3
The success of DNNs is driven by the counter-intuitive ability of over-parameterized networks to generalize, even when they perfectly fit the training data. In practice, test error…