Paper-Conference

WebMMU: A Benchmark for Multimodal Multilingual Website Understanding and Code Generation

VCR: Pixel-Level Complex Reasoning by Restoring Occluded Text

Rethinking Decentralized Learning: Towards More Realistic Evaluations with a Metadata-Agnostic Approach

MAP: Low-compute Model Merging with Amortized Pareto Fronts via Quadratic Approximation

BigDocs: An Open Dataset for Training Multimodal Models on Document and Code Tasks

Resonance RoPE: Improving Context Length Generalization of Large Language Models

Efficient Classification of Long Documents via State-Space Models

DiffTune: A Diffusion-Based Approach to Diverse Instruction-Tuning Data Generation

QEN: Applicable Taxonomy Completion via Evaluating Full Taxonomic Relations

Enquire One's Parent and Child Before Decision: Fully Exploit Hierarchical Structure for Self-Supervised Taxonomy Expansion