Model Mismatch

Model mismatch, the discrepancy between a model's assumptions and the real-world data it encounters, is a critical challenge across diverse machine learning applications. Current research focuses on mitigating this mismatch in various contexts, including reinforcement learning (through techniques like causal representation learning and robust optimization), graph neural networks (analyzing generalization under differing data generating processes), and inverse problems (using untrained neural network components to adapt to inaccurate forward models). Addressing model mismatch is crucial for improving the reliability and robustness of machine learning systems, leading to more accurate predictions and more effective decision-making in real-world scenarios.

Papers

November 12, 2024

Singularity-Avoidance Control of Robotic Systems with Model Mismatch and Actuator Constraints
Mingkun Wu, Alisa Rupenyan, Burkhard Corves
Learning Based Robotic System Control Barrier Function Control Constraint Model Mismatch Planar Robot Singularity Avoidance Kinematic Error

October 31, 2024

Exploring the Knowledge Mismatch Hypothesis: Hallucination Propensity in Small Models Fine-tuned on Data from Larger Models
Phil Wee, Riyadh Baghdadi
Large Language Model Fine Tuning Raw Data Large Model Content Hallucination Small Model Model Mismatch

September 10, 2024

One-Shot Imitation under Mismatched Execution
Kushal Kedia, Prithwish Dan, Angela Chao, Maximus Adrian Pace, Sanjiban Choudhury
Feature Imitation Long Horizon Human Demonstration Long Horizon Manipulation Task Cross Embodiment Model Mismatch

August 25, 2024

Generalization of Graph Neural Networks is Robust to Model Mismatch
Zhiyang Wang, Juan Cervino, Alejandro Ribeiro
Graph Neural Network Strong Generalization Generalization Gap Training Graph Model Mismatch Robust Graph Neural Network

August 24, 2024

Understanding Uncertainty-based Active Learning Under Model Mismatch
Amir Hossein Rahmati, Mingzhou Fan, Ruida Zhou, Nathan M. Urban, Byung-Jun Yoon, Xiaoning Qian
Prediction Uncertainty Prediction Performance Model Mismatch Uncertainty Based Active Learning Model Capacity

July 15, 2024

BECAUSE: Bilinear Causal Representation for Generalizable Offline Model-based Reinforcement Learning
Haohong Lin, Wenhao Ding, Jian Chen, Laixi Shi, Jiacheng Zhu, Bo Li, Ding Zhao
Model Based Reinforcement Learning Policy Learning Causal Representation Model Mismatch

May 2, 2024

Constrained Reinforcement Learning Under Model Mismatch
Zhongchang Sun, Sihong He, Fei Miao, Shaofeng Zou
Model Uncertainty Robust Policy Constrained Reinforcement Learning Reward Maximization Model Mismatch

March 7, 2024

Solving Inverse Problems with Model Mismatch using Untrained Neural Networks within Model-based Architectures
Peimeng Guan, Naveed Iqbal, Mark A. Davenport, Mudassir Masood
Inverse Problem Model Architecture Unlearned Model Deep Equilibrium Model Model Based Deep Learning Untrained Neural Network Model Mismatch Nonlinear Inverse Problem

January 1, 2024

Backdoor Attack on Unpaired Medical Image-Text Foundation Models: A Pilot Study on MedCLIP
Ruinan Jin, Chun-Yin Huang, Chenyu You, Xiaoxiao Li
Backdoor Attack Pilot Study Model Mismatch

December 5, 2023

Diffusion-Based Speech Enhancement in Matched and Mismatched Conditions Using a Heun-Based Sampler
Philippe Gonzalez, Zheng-Hua Tan, Jan Østergaard, Jesper Jensen, Tommy Sonne Alstrøm, Tobias May
Diffusion Model Generative Model Speech Enhancement Matching Accuracy Discriminative Model Model Mismatch Diffusion Based Speech Enhancement Reverse Sampler

October 31, 2023

The Alignment Ceiling: Objective Mismatch in Reinforcement Learning from Human Feedback
Nathan Lambert, Roberto Calandra
Reinforcement Learning Human Feedback Model Based Reinforcement Learning Reward Model Human Alignment Model Mismatch Reward Model Training

October 10, 2023

A Unified View on Solving Objective Mismatch in Model-Based Reinforcement Learning
Ran Wei, Nathan Lambert, Anthony McDonald, Alfredo Garcia, Roberto Calandra
Model Based Reinforcement Learning Policy OpTimization Unified View Learned Dynamic Reward Structure Model Mismatch

June 18, 2023

MISMATCH: Fine-grained Evaluation of Machine-generated Text with Mismatch Error Types
Keerthiram Murugesan, Sarathkrishna Swaminathan, Soham Dan, Subhajit Chaudhury, Chulaka Gunasekara, Maxwell Crouse, Diwakar Mahajan, Ibrahim Abdelaziz, Achille Fokoue, Pavan Kapanipathi, Salim Roukos, Alexander Gray
Large Language Model Fine Grained NLP Task Machine Generated Model Mismatch

November 16, 2022

An Empirical Analysis of the Advantages of Finite- v.s. Infinite-Width Bayesian Neural Networks
Jiayu Yao, Yaniv Yacoby, Beau Coker, Weiwei Pan, Finale Doshi-Velez
Scientific Inference Bayesian Neural Network Empirical Analysis Advantage Feedback Model Mismatch

October 12, 2022

Reducing The Mismatch Between Marginal and Learned Distributions in Neural Video Compression
Muhammet Balcilar, Bharath Bhushan Damodaran, Pierre Hellier
Image Compression Compression Technique Video Codecs Neural Video Compression Model Mismatch Trainable Layer Learning Probability Distribution Learning Based Video Compression

June 28, 2022

Volume-Independent Music Matching by Frequency Spectrum Comparison
Anthony Lee
Automatic Volume Model Mismatch Frequency Spectrum

April 10, 2022

Regret Analysis of Online Gradient Descent-based Iterative Learning Control with Model Mismatch
Efe C. Balta, Andrea Iannelli, Roy S. Smith, John Lygeros
Online Learning Regret Analysis Gradient Information Online Gradient Descent Feedforward Control Model Mismatch Iterative Learning Control

March 7, 2022

December 10, 2021

PACMAN: PAC-style bounds accounting for the Mismatch between Accuracy and Negative log-loss
Matias Vera, Leonardo Rey Vega, Pablo Piantanida
Loss Function Generalization Error Generalization Gap Generalization Problem Model Mismatch Logarithmic Loss Negative Preference

Model Mismatch

Papers

Singularity-Avoidance Control of Robotic Systems with Model Mismatch and Actuator Constraints

Exploring the Knowledge Mismatch Hypothesis: Hallucination Propensity in Small Models Fine-tuned on Data from Larger Models

One-Shot Imitation under Mismatched Execution

Generalization of Graph Neural Networks is Robust to Model Mismatch

Understanding Uncertainty-based Active Learning Under Model Mismatch

BECAUSE: Bilinear Causal Representation for Generalizable Offline Model-based Reinforcement Learning

Constrained Reinforcement Learning Under Model Mismatch

Solving Inverse Problems with Model Mismatch using Untrained Neural Networks within Model-based Architectures

Backdoor Attack on Unpaired Medical Image-Text Foundation Models: A Pilot Study on MedCLIP

Diffusion-Based Speech Enhancement in Matched and Mismatched Conditions Using a Heun-Based Sampler

The Alignment Ceiling: Objective Mismatch in Reinforcement Learning from Human Feedback

A Unified View on Solving Objective Mismatch in Model-Based Reinforcement Learning

MISMATCH: Fine-grained Evaluation of Machine-generated Text with Mismatch Error Types

An Empirical Analysis of the Advantages of Finite- v.s. Infinite-Width Bayesian Neural Networks

Reducing The Mismatch Between Marginal and Learned Distributions in Neural Video Compression

Volume-Independent Music Matching by Frequency Spectrum Comparison

Regret Analysis of Online Gradient Descent-based Iterative Learning Control with Model Mismatch

Estimation under Model Misspecification with Fake Features

Mismatch between Multi-turn Dialogue and its Evaluation Metric in Dialogue State Tracking

PACMAN: PAC-style bounds accounting for the Mismatch between Accuracy and Negative log-loss