arXiv:2510. 04212v4 Announce Type: replace-cross Abstract: The pursuit of computational efficiency has driven the adoption of low-precision formats for training transformer models.
arXiv:2505. 15548v2 Announce Type: replace Abstract: Autoregressive transformer language models frequently exhibit training instability when trained on long sequences, particularly under low-precision arithmetic.
arXiv:2607. 09800v1 Announce Type: new Abstract: Training in reduced floating-point precision can silently halt learning: when a gradient-descent weight update falls below half the unit in the last place (ULP) of the weight, it rounds away and that coordinate freezes while its gradient is still nonzero.
arXiv:2505. 01043v2 Announce Type: replace Abstract: Large language models (LLMs) have achieved impressive performance across various domains.
arXiv:2604. 13082v2 Announce Type: replace-cross Abstract: Grokking in transformers trained on algorithmic tasks is characterized by a long delay between training-set fit and abrupt generalization, but the source of that delay remains poorly understood.
arXiv:2410. 24050v3 Announce Type: replace Abstract: Large-scale pretraining of transformers has been central to the success of foundation models.
arXiv:2604. 10098v2 Announce Type: replace Abstract: As the foundational architecture of modern machine learning, Transformers have driven remarkable progress across diverse AI domains.
arXiv:2608. 07921v1 Announce Type: cross Abstract: We apply Marchenko-Pastur (MP) random matrix theory to pre-trained attention weights in order to separate each projection matrix into a random-like bulk and a set of spectral outliers.
arXiv:2502. 08363v3 Announce Type: replace-cross Abstract: We present Top-Theta (Top-$\theta$) Attention, a training-free method for sparsifying transformer attention during inference.
arXiv:2603. 00040v3 Announce Type: replace-cross Abstract: Achieving reliable 4-bit attention is a prerequisite for end-to-end FP4 computation on emerging FP4-capable GPUs, yet attention remains the main obstacle due to FP4's tiny dynamic range and attention's heavy-tailed activations.
arXiv:2603. 06591v2 Announce Type: replace Abstract: Transformers frequently allocate disproportionate attention to specific tokens, a phenomenon known as attention sinks.
arXiv:2606. 27449v1 Announce Type: new Abstract: Multi-head attention conventionally partitions the hidden dimension equally across all heads at every layer, enforcing an identical representational subspace dimension (dh = dmodel/h) throughout the models depth.
arXiv:2605. 09825v4 Announce Type: replace-cross Abstract: Why does full-pipeline FP4 training of large language models often diverge, even when forward activations and activation gradients remain stable?