paper-with-me

홈 › Papers

Massive Spikes in LLMs are Bias Vectors: Mechanistic Uncovering and Spike-Free Quantization

2026-06-01 · Yung-Chin Chen, Chung Peng Lee, Ze-Wei Liou, Naveen Verma arxiv

Massive activation spikes in Large Language Models (LLMs) severely degrade quantization by stretching dynamic ranges. While prior hypotheses characterize these as high-level scalar biases, we argue that they are merely the scalar intermediates of rigid, structural vector biases in the spike-carrying tokens. We show that these tokens converge to constant vectors after normalization that drive the attention sink and value-state drain mechanisms. We geometrically substantiate this by analyzing the coordination of projection weights: $W_K$ contrastively amplifies the vector, $W_Q$ aligns semantic tokens toward it, and $W_V$ projects it into the spectral null-space. Furthermore, we reveal that the model actively preserves these structural biases against Rotary Positional Embedding (RoPE) perturbations by localizing them in "zones of rotational stability" utilizing low-frequency bands and coherent channel pairs. Leveraging this, we propose INSERTQUANT, a post-training quantization (PTQ) framework that clamps spikes and restores their function via pre-computed template vectors. This renders activations strictly spike-free, enabling robust low-bit quantization with high fidelity. INSERTQUANT achieves parity with state-of-the-art per-tensor quantization methods on LLMs and uniquely generalizes beyond text to other modalities such as ViTs.

📄 PDF Abstract BibTeX arXiv:2606.02288

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Massive Activations in Hybrid Linear Attention Large Language Models: Pre-Attention Spikes and Inter-Spike Plateaus

2026-08-12 · Zunhai Su, Bohan Sun, Xialie Zhuang, Shuibai Zhang 외 hf

We present the first systematic study of Massive activations (MAs) in layer-interleaved HLA LLMs and uncover two architecture-aligned morphologies: MAs consistently spike immediately before full attention layers, forming…

Anchored Answers: Unravelling Positional Bias in GPT-2's Multiple-Choice Questions

2024-05-06 · Ruizhe Li, Yanjun Gao

Large Language Models (LLMs), such as the GPT-4 and LLaMA families, have demonstrated considerable success across diverse tasks, including multiple-choice questions (MCQs). However, these models exhibit a positional bias…

Decision MakingMultiple-choice

Activation Steering for Bias Mitigation: An Interpretable Approach to Safer LLMs

2025-08-12 · Shivam Dubey arxiv

As large language models (LLMs) become more integrated into societal systems, the risk of them perpetuating and amplifying harmful biases becomes a critical safety concern. Traditional methods for mitigating bias often r…

Small Vectors, Big Effects: A Mechanistic Study of RL-Induced Reasoning via Steering Vectors

2025-09-08 · Viacheslav Sinii, Nikita Balagansky, Gleb Gerasimov, Daniil Laptev 외 arxiv

The mechanisms by which reasoning training reshapes LLMs' internal computations remain unclear. We study lightweight steering vectors inserted into the base model's residual stream and trained with a reinforcement-learni…

LLM Assertiveness can be Mechanistically Decomposed into Emotional and Logical Components

2025-08-24 · Hikaru Tsujimura, Arush Tagade arxiv

Large Language Models (LLMs) often display overconfidence, presenting information with unwarranted certainty in high-stakes contexts. We investigate the internal basis of this behavior via mechanistic interpretability. U…