Influence functions are inaccurate in deep learning models, especially for deeper networks.
problem Inaccuracies in influence functions in deep learning models.
method Empirical study of influence functions in neural network models trained on various datasets.
result Influence estimates are often erroneous for deeper networks and require regularization.
RelatIF selects more intuitive training examples for explaining model predictions.
problem Influence functions identify outliers as explanatory examples, leading to poor explanations.
method RelatIF separates global and local influence, optimizing for local relative to global effects.
result Examples selected by RelatIF are more intuitive than those from influence functions.
Establishes statistical and computational bounds for influence diagnostics.
problem Identifying influential datapoints or subsets in machine learning models.
method Finite-sample statistical bounds and computational complexity for influence functions and approximate maximum influence perturbations.
result Established statistical and computational guarantees for influence diagnostics.
Influence functions help study large language model generalization, revealing surprising decay patterns.
problem Understanding and mitigating risks in large language models (LLMs).
method Eigenvalue-corrected Kronecker-Factored Approximation (EK-FAC) to scale influence functions to LLMs.
result Influences decay to near-zero when key phrases order is flipped, revealing a surprising limitation.
New research investigates why influence functions are fragile and proposes new validation procedures.
problem Understanding and mitigating the fragility of influence functions in deep learning model explanations.
method Verification of influence functions using various conditions and procedures, including convexity and non-convexity.
result Validation procedures may cause the observed fragility of influence functions.
Study shows influence functions are poor for neural networks but useful for identifying influential examples.
problem Influence functions misalign with leave-one-out retraining in neural networks.
method Decomposed the discrepancy into five terms and studied their contributions across different architectures and datasets.
result Influence functions are a good approximation to the proximal Bregman response function (PBRF), useful for identifying influential examples.
Paper introduces a multi-stage influence function to track model predictions.
problem Improving natural language processing and computer vision performance.
method Develops a multi-stage influence function to track predictions from finetuned models back to pretraining data.
result Identifies pretraining examples contributing most to finetuning task predictions.
The paper extends influence functions to sequence tagging tasks for better model interpretability.
problem Lack of interpretability methods for sequence tagging models.
method Define and compute influence of training instance segments on test segment predictions.
result The segment influence method tracks with true influence and identifies annotation errors.
Better Hessian approximations improve influence function attributions in deep learning.
problem Influence functions are difficult to compute due to ill-conditioned Hessians, leading to poor data attribution performance.
method Investigated the impact of Hessian approximation quality on influence-function attributions in a controlled setting.
result Better Hessian approximations consistently yield better influence score quality.
Enhances influence functions for deep models without costly Hessian inversion.
problem Challenges in computing Hessian matrix inverse for deep models.
method Substitutes Hessian inverse with identity matrix and extends applications.
result TracIn method performs well in measuring model utility and fairness.
The paper simplifies influence computations for large-scale machine learning models.
problem Improving training efficiency and accuracy in large-scale models.
method Study influence functions, define memorization, simplify computations.
result Influence functions can be practical for large-scale models, indicating memorization.
New IF method improves accuracy in deep neural networks with noisy data.
problem Inaccurate influence estimates in deep neural networks, especially with noisy data.
method Established a connection between influence estimation error, validation set risk, and sharpness, introducing a novel estimation form for flat validation minima.
result Our novel Influence Function approach provides more accurate influence estimates, validated across various tasks.
A new method, VIF, calculates influence for non-decomposable losses efficiently.
problem Efficiently calculating influence for complex machine learning models with non-decomposable losses.
method Revisiting influence function from robust statistics, proposing Versatile Influence Function (VIF) for any non-decomposable loss.
result VIF method is up to 10^3 times faster than brute-force methods and closely matches influence results.
Develops a faster model selection method using influence functions.
problem Time-consuming and computationally inefficient model selection.
method Model-specific data subsampling guided by influence functions.
result Empirically shows improved selection of high-quality models.
New algorithm for optimizing statistical utilities in bandits.
problem Optimizing statistical functionals of long-run reward distributions.
method Influence-function calculus for stochastic gradient estimation, entropic mirror-ascent algorithm.
result Regret bounds that separate optimization and estimation errors.
How can we explain the predictions of a black-box model? In this paper, we use influence functions -- a classic technique from robust statistics -- to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To …
Paper presents a new way to estimate model changes without full model evaluation.
problem Efficiently estimating changes in model parameters and outputs due to data point removal.
method Dual representation of influence functions for linearizable models, reducing computational complexity.
result The dual representation can be an efficient alternative to original influence functions, especially for large models.
With the rapid adoption of machine learning systems in sensitive applications, there is an increasing need to make black-box models explainable. Often we want to identify an influential group of training samples in a particular test prediction for a given machine learning model. Existing influence functions tackle this…
Improved scalability and interpretability in training data attribution.
problem Identifying which training data drives specific behaviors, especially unintended ones.
method Leveraging interpretable structures within the model to attribute model behavior to semantic directions, not individual test examples.
result Simple probe-based attribution methods are first-order approximations of Concept Influence that achieve comparable performance while being over an order-of-magnitude faster.
Estimates spatio-temporal Hawkes processes using tensor recovery.
problem Estimating influence functions for spatio-temporal Hawkes processes.
method Formulates influence function as a tensor kernel, assumes low-rank structure, solves as convex optimization problem.
result Provides theoretical guarantees and demonstrates efficiency with simulations.
Improves understanding of PWS by calculating influence of sources and data.
problem Understanding the influence of each component in PWS.
method Proposes source-aware Influence Function (IF) to decompose and calculate influence.
result Improves end model's generalization performance and identifies mislabeling.
Latent factor models (LFMs) such as matrix factorization achieve the state-of-the-art performance among various Collaborative Filtering (CF) approaches for recommendation. Despite the high recommendation accuracy of LFMs, a critical issue to be resolved is the lack of explainability. Extensive efforts have been made in…
DataInf efficiently approximates data influence in large models, improving transparency and identifying mislabeled data.
problem Efficiently estimating data influence in large-scale models like LoRA-tuned LLMs and diffusion models.
method DataInf uses a closed-form expression to approximate influence scores efficiently.
result DataInf outperforms existing methods in computational and memory efficiency, accurately identifying influential data points.
MISS identifies influential subsets in ML models.
problem Capturing collective influence of training samples.
method Analyzed and compared influence-based greedy heuristics and adaptive versions.
result Adaptive heuristics can better capture sample interactions.
An artificial general intelligence (AGI) might have an instrumental drive to modify its utility function to improve its ability to cooperate, bargain, promise, threaten, and resist and engage in blackmail. Such an AGI would necessarily have a utility function that was at least partially observable and that was influenc…
Influence functions estimate the effect of removing a training point on a model without the need to retrain. They are based on a first-order Taylor approximation that is guaranteed to be accurate for sufficiently small changes to the model, and so are commonly used to study the effect of individual points in large data…
Paper introduces Influence Function to assess OOD generalization stability.
problem Assessing OOD generalization accuracy when target domains are unknown.
method Introduced Influence Function from robust statistics to monitor model stability.
result Accuracy on test domains and Influence Function variance can distinguish OOD algorithms and generalization quality.
Newfluence improves model interpretability in high-dimensional AI models.
problem Challenges in interpreting high-dimensional AI models.
method Introduced Newfluence, an alternative approximation to influence functions.
result Newfluence offers significantly improved accuracy in high-dimensional settings.
Sub-sampling is a common and often effective method to deal with the computational challenges of large datasets. However, for most statistical models, there is no well-motivated approach for drawing a non-uniform subsample. We show that the concept of an asymptotically linear estimator and the associated influence func…
Improved data attribution for high-dimensional models.
problem Inaccurate influence functions in high-dimensional settings.
method Rescaled influence functions (RIF) for more precise data attribution.
result RIFs offer significantly better predictions in practice.
Proposes DeGLIF to denoise graph data for label noise robustness.
problem Label noise in graph data makes node classification challenging.
method Uses leave-one-out influence function to denoise graph data.
result DeGLIF improves accuracy in node classification on noisy datasets.
HyperINF improves influence function estimation for large models with better accuracy and efficiency.
problem Inaccurate and computationally expensive influence function estimation for large-scale models.
method HyperINF leverages Schulz's iterative algorithm and GFIM for low-rank approximation of Hessian matrix.
result HyperINF achieves superior accuracy and performance compared to existing methods on LoRA-tuned models.
Prediction-powered causal inference achieves smaller asymptotic variance than traditional methods.
problem Estimating causal and structural parameters in a semi-supervised setting.
method Combining efficient influence function with debiased machine learning and semi-supervised Riesz regression.
result Asymptotic variances of estimators match the derived efficiency bound.
In this paper, we study the problem of robust influence maximization in the independent cascade model under a hyperparametric assumption. In social networks users influence and are influenced by individuals with similar characteristics and as such, they are associated with some features. A recent surging research direc…
GOIMDA selects inputs to maximize expected influence on a goal functional, reducing data acquisition needs.
problem Challenges in active data acquisition for learning and optimization tasks in deep neural networks.
method GOIMDA uses inverse curvature and goal gradient to select inputs maximizing expected influence on a specified goal functional.
result GOIMDA achieves target performance with fewer labeled samples or function evaluations compared to baselines.
Paper introduces AIF to analyze robust optimization effects.
problem Quantifying robust optimization's impact on model optimizers and losses.
method Inspired by robust statistics, AIF is introduced to measure model sensitivity.
result AIF reveals how model complexity and randomized smoothing affect model sensitivity.
We optimize discounts to maximize influence spread in social networks.
problem Maximizing influence spread in social networks with fractional discounts.
method Developed an efficient (1-1/e)-approximation algorithm for NP-hard problem.
result Achieved an approximation of 1-1/e for influence maximization.
In2Core selects a coreset for efficient LLM fine-tuning with reduced data.
problem Costly fine-tuning of large language models due to extensive parameters and data requirements.
method Analyzes model gradients to estimate training sample influence, optimizing for efficiency.
result Achieves similar performance with 50% of training data using In2Core.
Captures data influence changes during training.
problem Traditional influence functions fail for modern training methods.
method Formalized trajectory-specific LOO influence, using data value embedding.
result Data influence varies by training stage, early and late stages have greater impact.
New stable HOIF estimators for statistical functionals.
problem Constructing numerically stable HOIF estimators for statistical functionals.
method Developed new sHOIF estimators with provable guarantees.
result 2nd order sHOIF estimators were validated in synthetic experiments.
Efficient estimators for smooth Hilbert-valued parameters with theoretical guarantees.
problem Estimating smooth Hilbert-valued parameters with theoretical guarantees.
method Pathwise differentiable Hilbert-valued parameters, efficient influence functions, regularized one-step estimators.
result Theoretical guarantees for efficient estimators even when nuisance functions are arbitrary.
Proposes a new estimator for causal mediation with continuous treatments.
problem Estimation of direct and indirect effects with continuous treatments.
method Kernel smoothing approach with cross-fitting for non-parametric estimation.
result Multiply robust and asymptotically normal estimator for continuous treatments.
Develops new methods to evaluate data influence in SAM for improved model training.
problem Challenges in mislabeled noisy data and privacy concerns in SAM.
method Two innovative data valuation methods based on influence functions (IF) for SAM.
result Demonstrates effectiveness in identifying mislabeled data and enhancing interpretability.
We study the problem of learning influence functions under incomplete observations of node activations. Incomplete observations are a major concern as most (online and real-world) social networks are not fully observable. We establish both proper and improper PAC learnability of influence functions under randomly missi…
We propose and analyze estimators for statistical functionals of one or more distributions under nonparametric assumptions. Our estimators are based on the theory of influence functions, which appear in the semiparametric statistics literature. We show that estimators based either on data-splitting or a leave-one-out t…
We consider a crowdsourcing data acquisition scenario, such as federated learning, where a Center collects data points from a set of rational Agents, with the aim of training a model. For linear regression models, we show how a payment structure can be designed to incentivize the agents to provide high-quality data as …
New method reveals true causal functions in nonlinear time series, not just scores.
problem Causal discovery in nonlinear time series often uses scalar edge scores, which hide true function-valued causal influence.
method Formalized function-valued causal influence for additive, contribution-decomposable architectures. Introduced a practical framework based on ICE for estimating causal response functions directly from trained models.
result Edges with indistinguishable scalar scores can exhibit qualitatively different functional behaviors.
New estimator stabilizes higher-order influence functions for bilinear forms.
problem Stability issues in estimating bilinear forms using higher-order influence functions.
method Proposes a new stabilized higher-order estimator for a class of bilinear forms without sample splitting.
result New estimator exhibits more stable finite-sample performance compared to the empirical higher-order estimator.