# A Unified View of Attention and Residual Sinks: Outlier-Driven Rescaling is Essential for Transformer Training

arXiv: https://arxiv.org/abs/2601.22966 · HTML: https://mlsanity.com/paper/2601.22966/

## Key facts

- Cited by 23 papers in the map (data through 14 Oct 2026).
- Builds on 53 of its 60 references that are in the map.
- First posted on arXiv 30 Jan 2026.

## Builds on (most cited first, 30 of 53)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (2021)
- [Program Synthesis with Large Language Models](https://mlsanity.com/paper/2108.07732/) (2021)
- [Analyzing and Improving the Image Quality of StyleGAN](https://mlsanity.com/paper/1912.04958/) (2019)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [FiLM: Visual Reasoning with a General Conditioning Layer](https://mlsanity.com/paper/1709.07871/) (2017)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://mlsanity.com/paper/1909.08053/) (2019)
- [MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark](https://mlsanity.com/paper/2406.01574/) (2024)
- [GLU Variants Improve Transformer](https://mlsanity.com/paper/2002.05202/) (2020)
- [Root Mean Square Layer Normalization](https://mlsanity.com/paper/1910.07467/) (2019)
- [SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models](https://mlsanity.com/paper/2211.10438/) (2022)
- [gpt-oss-120b & gpt-oss-20b Model Card](https://mlsanity.com/paper/2508.10925/) (2025)
- [Gated Delta Networks: Improving Mamba2 with Delta Rule](https://mlsanity.com/paper/2412.06464/) (2024)
- [Language Models are Multilingual Chain-of-Thought Reasoners](https://mlsanity.com/paper/2210.03057/) (2022)
- [ZeroQuant: Efficient and Affordable Post-Training Quantization for Large-Scale Transformers](https://mlsanity.com/paper/2206.01861/) (2022)
- [Analyzing and Improving the Training Dynamics of Diffusion Models](https://mlsanity.com/paper/2312.02696/) (2023)
- [CRUXEval: A Benchmark for Code Reasoning, Understanding and Execution](https://mlsanity.com/paper/2401.03065/) (2024)
- [Massive Activations in Large Language Models](https://mlsanity.com/paper/2402.17762/) (2024)
- [Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free](https://mlsanity.com/paper/2505.06708/) (2025)
- [Understanding and Improving Layer Normalization](https://mlsanity.com/paper/1911.07013/) (2019)
- [When Attention Sink Emerges in Language Models: An Empirical View](https://mlsanity.com/paper/2410.10781/) (2024)
- [Are We Done with MMLU?](https://mlsanity.com/paper/2406.04127/) (2024)
- [Outlier Suppression: Pushing the Limit of Low-bit Transformer Language Models](https://mlsanity.com/paper/2209.13325/) (2022)
- [Quantizable Transformers: Removing Outliers by Helping Attention Heads Do Nothing](https://mlsanity.com/paper/2306.12929/) (2023)
- [Understanding and Overcoming the Challenges of Efficient Transformer Quantization](https://mlsanity.com/paper/2109.12948/) (2021)

## Built on by (most cited first, 23 of 23)

- [On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability](https://mlsanity.com/paper/2608.30320/) (2026)
- [Rethinking Language Model Scaling under Transferable Hypersphere Optimization](https://mlsanity.com/paper/2603.28743/) (2026)
- [Attention Sinks Are Provably Necessary in Softmax Transformers: Evidence from Trigger-Conditional Tasks](https://mlsanity.com/paper/2603.11487/) (2026)
- [Attention Sink in Transformers: A Survey on Utilization, Interpretation, and Mitigation](https://mlsanity.com/paper/2604.10098/) (2026)
- [SiameseNorm: Breaking the Barrier to Reconciling Pre/Post-Norm](https://mlsanity.com/paper/2602.08064/) (2026)
- [Negligible in Size, Significant in Effect: On Scale Vectors in Large Language Models](https://mlsanity.com/paper/2605.26895/) (2026)
- [What Makes Position Zero Special? A Mechanistic Study of Position Zero Attention Sinks in LLMs](https://mlsanity.com/paper/2603.06591/) (2026)
- [Attention Sinks Induce Gradient Sinks: Massive Activations as Gradient Regulators in Transformers](https://mlsanity.com/paper/2603.17771/) (2026)
- [Steering Video Diffusion Transformers with Massive Activations](https://mlsanity.com/paper/2603.17825/) (2026)
- [On the Relationship Between Activation Outliers and Feature Death in Sparse Autoencoders](https://mlsanity.com/paper/2605.31518/) (2026)
- [Geometric Analysis of Token Selection in Multi-Head Attention](https://mlsanity.com/paper/2602.01893/) (2026)
- [Compact Task-Aligned Imitation Learning for Laboratory Automation](https://mlsanity.com/paper/2603.01110/) (2026)
- [Colinearity Decay: Training Quantization-Friendly ViTs with Outlier Decay](https://mlsanity.com/paper/2605.01330/) (2026)
- [Taming Outlier Tokens in Diffusion Transformers](https://mlsanity.com/paper/2605.05206/) (2026)
- [Sink vs. diagonal patterns as mechanisms for attention switch and oversmoothing prevention](https://mlsanity.com/paper/2605.08453/) (2026)
- [Hybrid Gated Attention](https://mlsanity.com/paper/2608.11805/) (2026)
- [A Mechanistic Account of Attention Sinks in GPT-2: One Circuit, Broader Implications for Mitigation](https://mlsanity.com/paper/2604.14722/) (2026)
- [Motif 3: Technical Report](https://mlsanity.com/paper/2608.09119/) (2026)
- [Cracks in the Foundation: Seemingly Minor Architectural Choices Impact Long Context Extension](https://mlsanity.com/paper/2608.10296/) (2026)
- [A.X K2 Technical Report](https://mlsanity.com/paper/2608.30181/) (2026)
- [Abstention and Noise Filtering: Two Missing Primitives of Softmax Attention](https://mlsanity.com/paper/2609.22005/) (2026)
- [Transformer MLP Gate Thresholds Are Couplings to a Carried Reference Direction](https://mlsanity.com/paper/2609.31956/) (2026)
- [IronLLM: Forging Compact Edge-Native Language Models for Real-Time Embodied Intelligence](https://mlsanity.com/paper/2609.36860/) (2026)

## Abstract (arXiv)

We investigate the functional role of emergent outliers in large language models, specifically attention sinks (a few tokens that consistently receive large attention logits) and residual sinks (a few fixed dimensions with persistently large activations across most tokens). We hypothesize that these outliers, in conjunction with the corresponding normalizations (textite.g., softmax attention and RMSNorm), effectively rescale other non-outlier components. We term this phenomenon textitoutlier-driven rescaling and validate this hypothesis across different model architectures and training token counts. This view unifies the origin and mitigation of both sink types. Our main conclusions and observations include: (1) Outliers function jointly with normalization: removing normalization eliminates the corresponding outliers but degrades training stability and performance; directly clipping outliers while retaining normalization leads to degradation, indicating that outlier-driven rescaling contributes to training stability. (2) Outliers serve more as rescale factors rather than contributors, as the final contributions of attention and residual sinks are significantly smaller than those of non-outliers. (3) Outliers can be absorbed into learnable parameters or mitigated via explicit gated rescaling, leading to improved training performance (average gain of 2 points) and enhanced quantization robustness (1.2 points degradation under W4A4 quantization).