Influence Functions
PulseAugur coverage of Influence Functions — every cluster mentioning Influence Functions across labs, papers, and developer communities, ranked by signal.
2 day(s) with sentiment data
-
New ZOIF method enables LLM data selection using only forward passes
Researchers have introduced ZOIF, a novel framework for estimating influence functions in large language models (LLMs) that bypasses the need for backpropagation. This method utilizes only forward passes, significantly …
-
New TICDA method attributes data influence in tabular foundation models
Researchers have developed TICDA, a novel method for attributing the influence of individual data points within the context of tabular foundation models (TFMs). Unlike existing methods that are computationally expensive…
-
New method enhances LLM training data attribution via response rewriting
Researchers have introduced a new method called influence-guided response rewriting to improve training data attribution (TDA) for large language models. This technique uses influence functions to identify influential t…
-
New method enhances language model control via influential training data
Researchers have developed a new method called influence-guided response rewriting to more effectively modify the behavior of language models. This technique identifies influential training data points using influence f…
-
New method characterizes tangent space for complex Markov models
Researchers have developed a method to characterize the orthogonal complement of the tangent space for general Markov models, a crucial step for efficient statistical inference. This work extends previous findings for d…
-
New theory extends statistical efficiency to Riemannian manifolds
A new paper by Lin Liu proposes an asymptotic efficiency theory applicable to statistical models with non-Euclidean structures, such as Riemannian manifolds. This work extends existing theories, which are largely confin…
-
New DRIFT method refines LLM training data for improved performance
Researchers have developed DRIFT, a novel method for refining instruction data to improve the performance ceiling of large language models. Unlike existing data curation techniques that focus on subset selection, DRIFT …
-
Influcoder offers scalable data attribution for LLMs
Researchers have developed Influcoder, a new method designed to efficiently attribute the influence of individual training data samples on large language models (LLMs). This approach addresses the scalability and speed …
-
New framework PINNfluence aids interpretation of physics-informed neural networks
Researchers have developed PINNfluence, a new framework designed to interpret the behavior of physics-informed neural networks (PINNs). This method utilizes influence functions to attribute the network's predictions and…
-
New CLIF method enhances NLP model interpretability with concept-level influence functions
Researchers have developed CLIF, a new method using influence functions to improve the interpretability of NLP models. This approach can identify influential training data points, both beneficial and detrimental, and al…
-
New analysis reveals accuracy of AI data attribution methods
Researchers have developed a new mathematical analysis for data attribution methods like Influence Functions (IF) and Newton Step (NS) in convex learning problems. This analysis does not rely on strong convexity assumpt…
-
New dual representation for influence functions improves efficiency
Researchers have developed a new dual representation for influence functions, which can efficiently estimate changes in model parameters and outputs. This method scales with dataset size rather than model size, offering…