accuracy improvement
Enhancements made to reduce errors in an AI model's predictions, often through algorithmic adjustments, additional training data, or optimization techniques.
- Accurate and Efficient Low-Rank Model Merging in Core Space
- Achieving $\tilde{\mathcal{O}}(1/N)$ Optimality Gap in Restless Bandits through Gaussian Approximation
- Asymmetric Duos: Sidekicks Improve Uncertainty
- BEDLAM2.0: Synthetic humans and cameras in motion
- BEDLAM2.0: Synthetic humans and cameras in motion
- Better Training Data Attribution via Better Inverse Hessian-Vector Products
- Compact Memory for Continual Logistic Regression
- Continual Release Moment Estimation with Differential Privacy
- Cross City Traffic Flow Generation via Retrieval Augmented Diffusion Model
- Cypher-RI: Reinforcement Learning for Integrating Schema Selection into Cypher Generation
- Dimension-free Score Matching and Time Bootstrapping for Diffusion Models
- DuoGPT: Training-free Dual Sparsity through Activation-aware Pruning in LLMs
- Enhancing the Outcome Reward-based RL Training of MLLMs with Self-Consistency Sampling
- Failure by Interference: Language Models Make Balanced Parentheses Errors When Faulty Mechanisms Overshadow Sound Ones
- Feature-Based Instance Neighbor Discovery: Advanced Stable Test-Time Adaptation in Dynamic World
- Frequency-Aware Token Reduction for Efficient Vision Transformer
- HeroFilter: Adaptive Spectral Graph Filter for Varying Heterophilic Relations
- Improving the Straight-Through Estimator with Zeroth-Order Information
- Language Models (Mostly) Know When to Stop Reading
- Learning to Specialize: Joint Gating-Expert Training for Adaptive MoEs in Decentralized Settings
- LogicTree: Improving Complex Reasoning of LLMs via Instantiated Multi-step Synthetic Logical Data
- MR. Video: MapReduce as an Effective Principle for Long Video Understanding
- Majority of the Bests: Improving Best-of-N via Bootstrapping
- Merging on the Fly Without Retraining: A Sequential Approach to Scalable Continual Model Merging
- Mysteries of the Deep: Role of Intermediate Representations in Out of Distribution Detection
- On the Role of Hidden States of Modern Hopfield Network in Transformer
- Optimizing Distributional Geometry Alignment with Optimal Transport for Generative Dataset Distillation
- PRESCRIBE: Predicting Single-Cell Responses with Bayesian Estimation
- Physics-informed machine learning with domain decomposition and global dynamics for three-dimensional intersecting flows
- Point-RFT: Improving Multimodal Reasoning with Visually Grounded Reinforcement Finetuning
- Progressive Data Dropout: An Embarrassingly Simple Approach to Train Faster
- QSVD: Efficient Low-rank Approximation for Unified Query-Key-Value Weight Compression in Low-Precision Vision-Language Models
- Rescaled Influence Functions: Accurate Data Attribution in High Dimension
- Right Question is Already Half the Answer: Fully Unsupervised LLM Reasoning Incentivization
- Rising from Ashes: Generalized Federated Learning via Dynamic Parameter Reset
- S-GRPO: Early Exit via Reinforcement Learning in Reasoning Models
- Sampling-Efficient Test-Time Scaling: Self-Estimating the Best-of-N Sampling in Early Decoding
- Scalable Evaluation and Neural Models for Compositional Generalization
- SpecReason: Fast and Accurate Inference-Time Compute via Speculative Reasoning
- Temporal Chain of Thought: Long-Video Understanding by Thinking in Frames
- Thinker: Learning to Think Fast and Slow
- Thoughts Are All Over the Place: On the Underthinking of Long Reasoning Models
- UniHG: A Large-scale Universal Heterogeneous Graph Dataset and Benchmark for Representation Learning and Cross-Domain Transferring
- UtilGen: Utility-Centric Generative Data Augmentation with Dual-Level Task Adaptation
- VeriThinker: Learning to Verify Makes Reasoning Model Efficient
- Web-Shepherd: Advancing PRMs for Reinforcing Web Agents