Mathematical Derivation Graphs: A Task for Summarizing Equation Dependencies in STEM Manuscriptsby Vishesh Prasad, Brian…
HoPE: A Novel Positional Encoding Without Long-Term Decay for Enhanced Context Awareness and Extrapolationby Yuhan…
LongReward: Improving Long-context Large Language Models with AI Feedbackby Jiajie Zhang, Zhongni Hou, Xin Lv,…
Layer by Layer: Uncovering Where Multi-Task Learning Happens in Instruction-Tuned Large Language Modelsby Zheng Zhao,…
Notes on the Mathematical Structure of GPT LLM Architecturesby Spencer Becker-KahnFirst submitted to arxiv on:…
Indication Finding: a novel use case for representation learningby Maren Eckhoff, Valmir Selimi, Alexander Aranovitch,…
TesseraQ: Ultra Low-Bit LLM Post-Training Quantization with Block Reconstructionby Yuhang Li, Priyadarshini PandaFirst submitted to…
Rethinking Positive Pairs in Contrastive Learningby Jiantao Wu, Shentong Mo, Zhenhua Feng, Sara Atito, Josef…
PETAH: Parameter Efficient Task Adaptation for Hybrid Transformers in a resource-limited Contextby Maximilian Augustin, Syed…
LiNeS: Post-training Layer Scaling Prevents Forgetting and Enhances Model Mergingby Ke Wang, Nikolaos Dimitriadis, Alessandro…