Add literature review protocol, searches and first snowball round

Six seeds snowballed (Semantic Scholar) plus logged keyword searches
(OpenAlex, arXiv): 2,127 unique records, 480 title/abstract screened.
AI first-pass suggestions are in their own column; decisions are pending.

Claude-Session: https://claude.ai/code/session_01S8KN7NuUCm8nXLmy6eDuVv

02-literature/search/results/S2-openalex-logit-instability.csv

@@ -0,0 +1,12 @@
+source_id,arxiv,doi,year,title,first_author,n_authors,venue,citations,abstract
+W4416789464,10.48550/arxiv.2511.21377,10.48550/arxiv.2511.21377,2025,Controlling changes to attention logits,Ben Anson,2,arXiv (Cornell University),0,"Stability of neural network weights is critical when training transformer models. The query and key weights are particularly problematic, as they tend to grow large without any intervention. Applying normalization to queries and keys, known as `QK norm', fixes stability issues in practice, but is not always applicable. For example, QK norm is not compatible with Multi Latent Attention (MLA) because QK norm requires full materialization of queries and keys during inference, which is not done in MLA. In this paper we suggest that controlling the changes to logits is important for stability. We show that these changes are controllable by assigning parameter-dependent learning rates to the query and key weights. We find that our cheap intervention allows us to increase the base learning rate of the network, outperform other methods in the MLA setting, and achieve performance competitive with QK norm when using Multi-head Attention."
+W4415329074,10.48550/arxiv.2505.15548,10.48550/arxiv.2505.15548,2025,Dense Local Dependencies Induce Attention-Logit Explosion and Training Instability During Long-Sequence Transformer Training,Suvadeep Hajra,1,arXiv (Cornell University),0,"Autoregressive transformer language models frequently exhibit training instability when trained on long sequences, particularly under low-precision arithmetic. Although this instability is often accompanied by attention-logit explosion, its underlying cause remains poorly understood. In this work, we present analytical insights and empirical evidence that dense local dependencies are a major contributor to attention-logit explosion. We demonstrate that dense local dependency patterns yield an effectively high-rank attention structure, which the low-rank parameterization of self-attention can only approximate with increasingly large logits as the sequence length grows. This logit inflation ultimately leads to training instability under low-precision arithmetic. We support this explanation through extensive experiments on synthetic and language modeling tasks. Our results consistently show that attention-logit growth increases with sequence length, is mitigated by increasing the attention dimension, and is substantially reduced by explicitly modeling dense local dependencies. Furthermore, we show that this growth is driven by the density of local dependencies rather than by locality alone. More broadly, our findings suggest that explicitly modeling dense local dependencies constitutes an important design principle for developing stable, efficient, and scalable long-context transformer architectures for autoregressive language modeling."
+W4324311371,10.48550/arxiv.2303.06296,10.48550/arxiv.2303.06296,2023,Stabilizing Transformer Training by Preventing Attention Entropy Collapse,Shuangfei Zhai,8,arXiv (Cornell University),13,"Training stability is of great importance to Transformers. In this work, we investigate the training dynamics of Transformers by examining the evolution of the attention layers. In particular, we track the attention entropy for each attention head during the course of training, which is a proxy for model sharpness. We identify a common pattern across different architectures and tasks, where low attention entropy is accompanied by high training instability, which can take the form of oscillating loss or divergence. We denote the pathologically low attention entropy, corresponding to highly concentrated attention scores, as $\textit{entropy collapse}$. As a remedy, we propose $σ$Reparam, a simple and efficient solution where we reparametrize all linear layers with spectral normalization and an additional learned scalar. We demonstrate that $σ$Reparam successfully prevents entropy collapse in the attention layers, promoting more stable training. Additionally, we prove a tight lower bound of the attention entropy, which decreases exponentially fast with the spectral norm of the attention logits, providing additional motivation for our approach. We conduct experiments with $σ$Reparam on image classification, image self-supervised learning, machine translation, speech recognition, and language modeling tasks. We show that $σ$Reparam provides stability and robustness with respect to the choice of hyperparameters, going so far as enabling training (a) a Vision Transformer {to competitive performance} without warmup, weight decay, layer normalization or adaptive optimizers; (b) deep architectures in machine translation and (c) speech recognition to competitive performance without warmup and adaptive optimizers. Code is available at \url{https://github.com/apple/ml-sigma-reparam}."
+W7106796243,,10.48448/nht8-e622,2025,Variance Sensitivity Induces Attention Entropy Collapse and Instability in Transformers,Association for Computational Linguistics 2025,3,Underline Science Inc.,0,"Attention-based language models typically rely on the softmax function to convert attention logits into probability distributions. However, this mechanism can lead to attention entropy collapse, where attention is focused on a single token, causing training instability. In this work, we identify the high variance sensitivity of softmax as a primary cause of this collapse. We show that entropy-stable attention mechanisms, which either control or are insensitive to the variance of attention logits, can prevent entropy collapse and enable more stable training. We provide empirical evidence of this effect in both large language models (LLMs) and a small Transformer model composed solely of self-attention and support our findings with theoretical analysis. Moreover, we identify that the concentration of attention probabilities increases the probability matrix norm, leading to a gradient exploding."
+W7212276107,10.48550/arxiv.2609.11655,10.48550/arxiv.2609.11655,2026,Musec: MomentUm SpEctral Clipping for Stable Muon-type Training,Zhuanghua Liu,3,arXiv (Cornell University),0,"Muon has emerged as a highly effective optimizer for large language model training, often achieving superior convergence and performance compared with the widely adopted Adam and AdamW optimizers. Nevertheless, Muon is prone to training instability due to its spectral flattening, manifested by loss spikes and unbounded growth of model weights. Existing approaches primarily rely on weight or attention-logit clipping, which require architecture-specific modifications and do not directly address instability across all model components. We propose MomentUm SpEctral Clipping (Musec), which replaces Muon's spectral flattening with spectral clipping: rather than setting all singular values of the momentum matrix to approximately one, Musec clips singular values that exceed a threshold while preserving the underlying spectral structure of the momentum. Our strategy provides an optimizer-level, architecture-agnostic mechanism for stabilizing Muon training. Theoretically, we establish convergence guarantees for Musec in nonconvex nonsmooth stochastic optimization. Practically, we develop Soft Musec, an efficient implementation that uses a smooth spectral saturation function approximated by coupled Newton-Schulz iterations. Empirically, Soft Musec consistently improves training stability over existing Muon variants across a wide range of learning rates and model sizes, remaining stable in settings where existing Muon variants diverge while matching their performance under well-tuned configurations."
+W7148294612,10.48550/arxiv.2604.00726,10.48550/arxiv.2604.00726,2026,Exploring Silent Data Corruption as a Reliability Challenge in LLM Training,Anton Altenbernd,3,arXiv (Cornell University),0,"As Large Language Models (LLMs) scale in size and complexity, the consequences of failures during training become increasingly severe. A major challenge arises from Silent Data Corruption (SDC): hardware-induced faults that bypass system-level detection mechanisms. SDC may behave like benign numerical noise, but can also cause harmful gradient corruption that leads to loss spikes, divergence, or stalled progress. This work provides a controlled study of how intermittent SDC affects LLM pretraining. Using targeted fault injection at the level of GPU matrix-multiply instructions, we characterize the sensitivity of different bit positions, kernel functions, and execution stages. Our analysis shows that locally originating faults can produce impactful corruption, including NaN propagation, short-lived spikes in loss, gradient norm, and attention logits, as well as persistent parameter divergence. Building on the observed corruption signatures, we propose a lightweight detection method that identifies potentially harmful parameter updates. Experiments on LLaMA models with 60M, 350M, and 1.3B parameters demonstrate that recomputing the most recent training step upon detection can effectively mitigate the impact of these events."
+W7149210485,,,2026,Exploring Silent Data Corruption as a Reliability Challenge in LLM Training,Anton Altenbernd,3,arXiv (Cornell University),0,"As Large Language Models (LLMs) scale in size and complexity, the consequences of failures during training become increasingly severe. A major challenge arises from Silent Data Corruption (SDC): hardware-induced faults that bypass system-level detection mechanisms. SDC may behave like benign numerical noise, but can also cause harmful gradient corruption that leads to loss spikes, divergence, or stalled progress. This work provides a controlled study of how intermittent SDC affects LLM pretraining. Using targeted fault injection at the level of GPU matrix-multiply instructions, we characterize the sensitivity of different bit positions, kernel functions, and execution stages. Our analysis shows that locally originating faults can produce impactful corruption, including NaN propagation, short-lived spikes in loss, gradient norm, and attention logits, as well as persistent parameter divergence. Building on the observed corruption signatures, we propose a lightweight detection method that identifies potentially harmful parameter updates. Experiments on LLaMA models with 60M, 350M, and 1.3B parameters demonstrate that recomputing the most recent training step upon detection can effectively mitigate the impact of these events."
+W4414448859,10.48550/arxiv.2508.20577,10.48550/arxiv.2508.20577,2025,MERIT: Maximum-normalized Element-wise Ratio for Language Model Large-batch Training,Yang Luo,6,arXiv (Cornell University),0,"Large-batch training has become a cornerstone in accelerating the training of deep neural networks, yet it poses challenges in optimization and generalization. Existing optimizers like AdamW present performance degradation during language models' large-batch training, due to the information bottleneck in attention layers caused by the sharp increase of max attention logit. While the LAMB optimizer partially addresses this issue, some attention layers still face this issue. The reason is that $l_2$-norm-based trust ratios in LAMB are less effective in directly influencing the max value of query/key weights. Furthermore, the weight-wise trust ratio in LAMB is error-prone as it overlooks relationships of weight values within rows or columns. Building on these observations, we propose a novel optimizer, MERIT, which leverages the max-norm to calculate the trust ratio to constrain the max attention logit more effectively. Moreover, we further construct element-wise trust ratios to provide more robust update scaling by focusing on local weight structures. Extensive experiments of large-batch training across various sizes of GPT-2 models demonstrate the superior performance of MERIT. Notably, during the training of GPT-2 Medium, MERIT enables a 6k batch size without any performance degradation compared to the standard batch size (480) with 48B training tokens. This work highlights the importance of considering the max attention logit and finer-granularity trust ratio in large-batch training. It successfully improves the training stability and paves the way for larger batch usage, enabling faster development and iteration of large language models. Code is available at https://github.com/NUS-HPC-AI-Lab/MERIT."
+W7127138976,10.48550/arxiv.2601.22966,10.48550/arxiv.2601.22966,2026,A Unified View of Attention and Residual Sinks: Outlier-Driven Rescaling is Essential for Transformer Training,Zihan Qiu,19,arXiv (Cornell University),0,"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 (\textit{e.g.}, softmax attention and RMSNorm), effectively rescale other non-outlier components. We term this phenomenon \textit{outlier-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)."
+W4414117222,,10.1109/fuzz62266.2025.11152151,2025,Fuzzy Confidence Attention: Harnessing Token-Level Uncertainty for Robust Transformer Models,Taniya Seth,2,IEEE International Conference on Fuzzy Systems (FUZZ-IEEE),0,"Transformer architectures have transformed natural language processing by capturing intricate dependencies through self-attention. Yet, conventional attention mechanisms lack a means to address token-level uncertainty, leaving models vulnerable to noisy, uncertain or unreliable inputs in real-world applications. In this paper, we propose fuzzy confidence attention—a novel mechanism that integrates token-level confidence estimates into the transformer’s attention framework via fuzzy sets. By mapping token confidence scores to fuzzy membership values using a differentiable, piecewise linear function, our method modulates attention logits to dynamically down-weight unreliable tokens. Experimental evaluations on controlled datasets reveal that our approach systematically reallocates attention, as shown by both qualitative heatmap visualizations and quantitative metrics such as Mean Noisy Token Attention and average KL divergence. These results demonstrate improved robustness and interpretability compared to standard attention mechanisms. FCA thus paves the way for more resilient transformer models in noisy and uncertain environments, with promising avenues for future work in large-scale and multimodal settings."
+W7161726329,10.48550/arxiv.2605.18106,10.48550/arxiv.2605.18106,2026,"Symmetry-Compatible Principle for Optimizer Design: Embeddings, LM Heads, SwiGLU MLPs, and MoE Routers",Tim Tsz-Kit Lau,2,arXiv (Cornell University),0,"A striking geometric disparity has long persisted in the practice of deep learning. While modern neural network architectures naturally exhibit rich symmetry and equivariance properties, popular optimizers such as Adam and its variants operate inherently coordinate-wise, rendering them unable to respect the equivariance structures of the parameter space. We address this disparity by introducing a symmetry-compatible principle for optimizer design: the gradient update rule should be equivariant under the symmetry group acting on the corresponding weight block. Following this principle, we first provide a unified perspective on bi-orthogonally equivariant updates for general matrix layers, as employed by stochastic spectral descent, Muon, Scion, and polar gradient methods. More importantly, by moving from orthogonal groups to permutation and shared-shift symmetries, we derive symmetry-compatible optimizers for parameter blocks whose symmetries differ from those of general matrix layers: embedding and LM head matrices, SwiGLU MLP projections, and MoE router matrices. These constructions include one-sided spectral, row-norm, hybrid row-norm/spectral, row-aware, column-aware, centered row-norm, and left-spectral updates. They yield an end-to-end layerwise optimizer stack in which each major matrix-valued parameter class is assigned an update whose equivariance matches its symmetry group. We corroborate this principle through pre-training experiments on dense and sparse MoE language models, including Qwen3-0.6B-style, Gemma 3 1B-style, OLMoE-1B-7B-style, and downsized gpt-oss architectures. Across these experiments, symmetry-compatible update rules consistently improve final validation loss, reduce expert load imbalance in sparse MoE models, and in several cases control final vocabulary-logit growth, improve router stability, and overall training stability over the corresponding AdamW updates."