Research Feed Page 42

HOMIE is a multimodal framework that improves how AI models generate personalized videos featuring consistent humans and objects by integrating multimodal guidance and identity-specific embeddings.

FlowMimic generates synthetic video editing data in real time by applying temporal flow fields to existing image editing samples, removing the need for labor-intensive mask annotations.

PPL-Factory is a method that selects the most informative subset of training data to improve fine-tuning performance while reducing computational overhead.

The paper introduces a causal inference framework to improve how RAG systems select actions by treating vector search as a nearest neighbor matching process.

The paper introduces a method that replaces compressed global visual vectors with uncompressed patch tokens to improve robotic manipulation precision while maintaining high computational efficiency.

The paper introduces a diagnostic benchmark that evaluates how well AI models simulate classical mechanics through a three-stage reasoning process.

The paper introduces a synthetic data generation framework that enables training API-calling agents without needing fully operational backend environments.

MagicSelector improves agent tool retrieval in mobile environments by using counterfactual reasoning and progressive reranking to eliminate semantic mismatches.

RynnBrain 1.1 introduces a unified framework that improves robotic manipulation by grounding visual understanding in physical space across different robot hardware.

The paper introduces an LLM-as-a-Coach method that replaces traditional scalar rewards in reinforcement learning with rich experiential knowledge to train open-ended task models.

Researchers improved Assamese speech recognition by fine-tuning the Whisper model using a combined dataset of validated and crowd-sourced audio.

The DADIR framework improves regression performance on imbalanced datasets by using adaptive data partitioning, density-regularized latent space learning, and targeted synthetic oversampling.

The authors introduce a set of differentially private algorithms that accelerate the max-sum diversification process while maintaining competitive utility.

The paper introduces mathematical frameworks and optimization algorithms to compute the Rate-Distortion-Perception function for better managing the trade-off between compression, accuracy, and perceptual quality.

The paper introduces the A-MESS framework to select and evaluate jailbreak attacks based on their actual contribution to improving model safety rather than just their success rate.

OrderMoE reduces cross-server communication overhead for Mixture of Experts models by grouping experts based on similarity and intelligently choosing between local and remote execution.

This paper investigates why large language models struggle with basic arithmetic and proposes human-inspired strategies, including task decomposition and specific prompting techniques, to improve their performance.

The paper introduces the first non-asymptotic analysis for the Navigate and Stop algorithm to identify optimal policies in online reinforcement learning with rigorous confidence bounds.

DynImmune-BERT introduces a continuous transformer architecture to predict cancer status by modeling longitudinal immune repertoire data as dynamic, event-driven processes.

The paper introduces a refined reinforcement learning method for LLM post-training that improves performance on math benchmarks by dynamically balancing teacher guidance and model autonomy.