Decoupling Dark Knowledge via Block-wise Logit Distillation for Feature-level Alignmentby Chengting Yu, Fengzhao Zhang, Ruizhe…
Methods of improving LLM training stabilityby Oleg Rybakov, Mike Chrzanowski, Peter Dykas, Jinze Xue, Ben…
Is Less More? Exploring Token Condensation as Training-free Test-time Adaptationby Zixin Wang, Dong Gong, Sen…
Estimating the Probabilities of Rare Outputs in Language Modelsby Gabriel Wu, Jacob HiltonFirst submitted to…
MLLM can see? Dynamic Correction Decoding for Hallucination Mitigationby Chenxi Wang, Xiang Chen, Ningyu Zhang,…
Divide, Reweight, and Conquer: A Logit Arithmetic Approach for In-Context Learningby Chengsong Huang, Langlin Huang,…
From Logits to Hierarchies: Hierarchical Clustering made Simpleby Emanuele Palumbo, Moritz Vandenhirtz, Alain Ryser, Imant…
Adjusting Pretrained Backbones for Performativityby Berker Demirel, Lingjing Kong, Kun Zhang, Theofanis Karaletsos, Celestine Mendler-Dünner,…
Towards Diverse Device Heterogeneous Federated Learning via Task Arithmetic Knowledge Integrationby Mahdi Morafah, Vyacheslav Kungurtsev,…
Embedding Geometries of Contrastive Language-Image Pre-Trainingby Jason Chuan-Chih Chou, Nahid AlamFirst submitted to arxiv on:…