Allowing each quantization group to select among multiple 4-bit grids improves accuracy over single-grid FP4 for both post-training and pre-training of LLMs.
Piqa: Reasoning about physical commonsense in natural language
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
citation-role summary
dataset 2
citation-polarity summary
years
2026 4roles
dataset 2polarities
use dataset 2representative citing papers
SKOP uses key-orthogonal projections to steer LLM activations while preserving attention patterns on focus tokens, cutting utility degradation by 5-7x and retaining over 95% of standard steering efficacy.
Different calibration objectives produce distinct layer pruning patterns in LLMs, while search algorithms converge to similar solutions under a fixed objective.
SMoA is a new PEFT adapter that uses block-wise Hadamard-modulated low-rank branches on spectral partitions to cover more pretrained spectral directions than standard LoRA under a smaller parameter budget.
citing papers explorer
-
Grid Games: The Power of Multiple Grids for Quantizing Large Language Models
Allowing each quantization group to select among multiple 4-bit grids improves accuracy over single-grid FP4 for both post-training and pre-training of LLMs.
-
Don't Lose Focus: Activation Steering via Key-Orthogonal Projections
SKOP uses key-orthogonal projections to steer LLM activations while preserving attention patterns on focus tokens, cutting utility degradation by 5-7x and retaining over 95% of standard steering efficacy.
-
Rethinking Layer Redundancy in Large Language Models: Calibration Objectives and Search for Depth Pruning
Different calibration objectives produce distinct layer pruning patterns in LLMs, while search algorithms converge to similar solutions under a fixed objective.
-
SMoA: Spectrum Modulation Adapter for Parameter-Efficient Fine-Tuning
SMoA is a new PEFT adapter that uses block-wise Hadamard-modulated low-rank branches on spectral partitions to cover more pretrained spectral directions than standard LoRA under a smaller parameter budget.