downstream tasks
Downstream tasks are activities or applications that utilize the results of pre-trained models. These tasks often involve fine-tuning or adapting models for specific use cases like classification, question answering, or other application-specific outputs.
- A Statistical Theory of Contrastive Learning via Approximate Sufficient Statistics
- Adaptive and Multi-scale Affinity Alignment for Hierarchical Contrastive Learning
- Advancing Expert Specialization for Better MoE
- CAT: Content-Adaptive Image Tokenization
- CellCLIP - Learning Perturbation Effects in Cell Painting via Text-Guided Contrastive Learning
- Closed-Form Training Dynamics Reveal Learned Features and Linear Structure in Word2Vec-like Models
- CoIDO: Efficient Data Selection for Visual Instruction Tuning via Coupled Importance-Diversity Optimization
- CodeMerge: Codebook-Guided Model Merging for Robust Test-Time Adaptation in Autonomous Driving
- Convergent Functions, Divergent Forms
- Cross-Domain Graph Data Scaling: A Showcase with Diffusion Models
- EgoDTM: Towards 3D-Aware Egocentric Video-Language Pretraining
- Enhancing 3D Reconstruction for Dynamic Scenes
- Filter Like You Test: Data-Driven Data Filtering for CLIP Pretraining
- FlexOLMo: Open Language Models for Flexible Data Use
- GRAVER: Generative Graph Vocabularies for Robust Graph Foundation Models Fine-tuning
- Geometry-Aware Edge Pooling for Graph Neural Networks
- GraphTOP: Graph Topology-Oriented Prompting for Graph Neural Networks
- Group-Level Data Selection for Efficient Pretraining
- HMVLM:Human Motion-Vision-Language Model via MoE LoRA
- Hardware-aligned Hierarchical Sparse Attention for Efficient Long-term Memory Access
- Implicit Modeling for Transferability Estimation of Vision Foundation Models
- LCDB 1.1: A Database Illustrating Learning Curves Are More Ill-Behaved Than Previously Thought
- LLM Generated Persona is a Promise with a Catch
- Latency NMS Attacks: Is It Real Life or Is It Just Fantasy?
- Learning Task-Agnostic Representations through Multi-Teacher Distillation
- LoRASuite: Efficient LoRA Adaptation Across Large Language Model Upgrades
- LoTA-QAF: Lossless Ternary Adaptation for Quantization-Aware Fine-Tuning
- Loquetier: A Virtualized Multi-LoRA Framework for Unified LLM Fine-tuning and Serving
- MMLongBench: Benchmarking Long-Context Vision-Language Models Effectively and Thoroughly
- MODEL SHAPLEY: Find Your Ideal Parameter Player via One Gradient Backpropagation
- Mars-Bench: A Benchmark for Evaluating Foundation Models for Mars Science Tasks
- MaxSup: Overcoming Representation Collapse in Label Smoothing
- Multi-Scale Finetuning for Encoder-based Time Series Foundation Models
- Multimodal 3D Genome Pre-training
- Native Segmentation Vision Transformers
- On the Loss of Context Awareness in General Instruction Fine-tuning
- On the Stability of Graph Convolutional Neural Networks: A Probabilistic Perspective
- OpenGU: A Comprehensive Benchmark for Graph Unlearning
- Orient Anything V2: Unifying Orientation and Rotation Understanding
- Orientation Matters: Making 3D Generative Models Orientation-Aligned
- PaZO: Preconditioned Accelerated Zeroth-Order Optimization for Fine-Tuning LLMs
- PathVQ: Reforming Computational Pathology Foundation Model for Whole Slide Image Analysis via Vector Quantization
- Periodic Skill Discovery
- Point-MaDi: Masked Autoencoding with Diffusion for Point Cloud Pre-training
- REVE: A Foundation Model for EEG - Adapting to Any Setup with Large-Scale Pretraining on 25,000 Subjects
- Reasoning is Periodicity? Improving Large Language Models Through Effective Periodicity Modeling
- Reward-oriented Causal Representation Learning
- SMARTraj$^2$: A Stable Multi-City Adaptive Method for Multi-View Spatio-Temporal Trajectory Representation Learning
- SPACE: Noise Contrastive Estimation Stabilizes Self-Play Fine-Tuning for Large Language Models
- Scalable Feature Learning on Huge Knowledge Graphs for Downstream Machine Learning
- Scaling Laws for Robust Comparison of Open Foundation Language-Vision Models and Datasets
- Simple and Effective Specialized Representations for Fair Classifiers
- Simultaneous Modeling of Protein Conformation and Dynamics via Autoregression
- Soft Task-Aware Routing of Experts for Equivariant Representation Learning
- Synthetic Series-Symbol Data Generation for Time Series Foundation Models
- The Impact of Coreset Selection on Spurious Correlations and Group Robustness
- The quest for the GRAph Level autoEncoder (GRALE)
- Towards A Translative Model of Sperm Whale Vocalization
- Towards Pre-trained Graph Condensation via Optimal Transport
- Towards Principled Unsupervised Multi-Agent Reinforcement Learning
- TrajMamba: An Efficient and Semantic-rich Vehicle Trajectory Pre-training Model
- Unleashing Foundation Vision Models: Adaptive Transfer for Diverse Data-Limited Scientific Domains
- VADB: A Large-Scale Video Aesthetic Database with Professional and Multi-Dimensional Annotations
- VIBE: Annotation-Free Video-to-Text Information Bottleneck Evaluation for TL;DR