Latest Articles for Data Evaluation

A study highlights data contamination's impact on code model evaluations.

A new dataset improves assessment of molecular knowledge in language models.

SPHINX-V enhances AI's ability to interpret images through user interaction.

BEAR improves assessment of relational knowledge in language models.

This study examines how language models handle different expressions of the same reasoning problems.

A new dataset evaluates how language models handle harmful content across cultures.

A new benchmark improves how we assess LVLMs and their accuracy.

An assessment of how well LLMs remember factual information and the factors involved.

This study offers improved methods for assessing text-to-image models.

A study evaluating few-shot learning methods for Polish language classification.

New metrics improve evaluation of information extraction systems in handwritten documents.

WorkBench tests agents' ability to perform realistic office tasks with a unique evaluation method.

Assessing how LLMs adapt to new information and biases.

A new method for assessing language models' alignment with human values.

Combining human reviewers with LLMs improves biomedical research evaluations.

A challenge focusing on deep generative models for realistic medical image generation.

A new system for assessing language models using real-world data streams.

A new method to assess commonsense reasoning in AI models through open-ended tasks.

New GAIA dataset sheds light on action quality in AI-generated content.

A new method to assess generative models with minimal data generation.

A new benchmark tests compositional reasoning in advanced models.

New dataset helps assess AI text accuracy and reliability.

A new benchmark evaluates how language models handle text changes.

A toolkit for assessing performance of retrieval-augmented models in specific domains.

VideoVista offers a comprehensive evaluation for video question-answering models.

Methods for measuring treatment effects across diverse groups and time frames.

This article presents a new method for assessing text-to-image models effectively.

Dysca introduces a new way to assess LVLM performance using synthetic data.

A new method measures how language models adapt their beliefs with new evidence.

A new benchmark to assess AI agents' performance in biomedical literature and knowledge graphs.

Introducing FairMedFM to evaluate the fairness of foundation models in healthcare.

This study assesses how medical LVLMs perform amidst hallucinations using a new dataset.

Exploring machine learning models and new datasets for improved security.

FKEA offers a fresh way to assess generative models without needing reference datasets.

A look at the benefits of segment-level evaluation methods for translation quality.

New metrics and EdgeHead module enhance 3D detection for autonomous vehicles.

A new approach enhances the accuracy of language model evaluations.

Improving how models handle evidence in long documents builds user trust.

BiasAlert enhances bias detection in language models for fairer AI outputs.

A new method to assess accuracy in language model outputs.