CoRect: Context-Aware Logit Contrast for Hidden State Rectification to Resolve Knowledge Conflicts
arXiv:2602.08221v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG) often struggles with knowledge conflicts, where model-internal parametric knowledge overrides retrieved evidence,…
Federated Prompt-Tuning with Heterogeneous and Incomplete Multimodal Client Data
arXiv:2602.07081v1 Announce Type: cross Abstract: This paper introduces a generalized federated prompt-tuning framework for practical scenarios where local datasets are…
SuReNav: Superpixel Graph-based Constraint Relaxation for Navigation in Over-constrained Environments
arXiv:2602.06807v1 Announce Type: cross Abstract: We address the over-constrained planning problem in semi-static environments. The planning objective is to find…
Towards Agentic Intelligence for Materials Science
arXiv:2602.00169v2 Announce Type: replace-cross Abstract: The convergence of artificial intelligence and materials science presents a transformative opportunity, but achieving true…
D-SCoRE: Document-Centric Segmentation and CoT Reasoning with Structured Export for QA-CoT Data Generation
arXiv:2508.01309v2 Announce Type: replace-cross Abstract: The scarcity and high cost of high-quality domain-specific question-answering (QA) datasets limit supervised fine-tuning of…
CSRv2: Unlocking Ultra-Sparse Embeddings
arXiv:2602.05735v2 Announce Type: replace-cross Abstract: In the era of large foundation models, the quality of embeddings has become a central…
Dr. Kernel: Reinforcement Learning Done Right for Triton Kernel Generations
arXiv:2602.05885v2 Announce Type: replace-cross Abstract: High-quality kernel is critical for scalable AI systems, and enabling LLMs to generate such code…
Supercharging Simulation-Based Inference for Bayesian Optimal Experimental Design
arXiv:2602.06900v1 Announce Type: cross Abstract: Bayesian optimal experimental design (BOED) seeks to maximize the expected information gain (EIG) of experiments.…
On the Identifiability of Steering Vectors in Large Language Models
arXiv:2602.06801v1 Announce Type: cross Abstract: Activation steering methods, such as persona vectors, are widely used to control large language model…
Emergent Low-Rank Training Dynamics in MLPs with Smooth Activations
arXiv:2602.06208v1 Announce Type: cross Abstract: Recent empirical evidence has demonstrated that the training dynamics of large-scale deep neural networks occur…