相关论文: Multipole Semantic Attention: A Fast Approximation…

The attention mechanism in Transformers is an important primitive for accurate and scalable sequence modeling. Its quadratic-compute and linear-memory complexity however remain significant bottlenecks. Linear attention and state-space…

Most language model pre-training frameworks concatenate multiple documents into fixed-length sequences and use causal masking to compute the likelihood of each token given its context; this strategy is widely adopted due to its simplicity…

Context-aware Machine Translation aims to improve translations of sentences by incorporating surrounding sentences as context. Towards this task, two main architectures have been applied, namely single-encoder (based on concatenation) and…

We propose Lizard, a linearization framework that transforms pretrained Transformer-based Large Language Models (LLMs) into subquadratic architectures. Transformers faces severe computational and memory bottlenecks with long sequences due…

As Large Language Models (LLMs) become increasingly prevalent in text simplification, systematically evaluating their outputs across diverse prompting strategies and architectures remains a critical methodological challenge in both NLP…

Linear transformers have emerged as a subquadratic-time alternative to softmax attention and have garnered significant interest due to their fixed-size recurrent state that lowers inference cost. However, their original formulation suffers…

Lexical ambiguity is a significant and pervasive challenge in Neural Machine Translation (NMT), with many state-of-the-art (SOTA) NMT systems struggling to handle polysemous words (Campolungo et al., 2022). The same holds for the NMT…

We present DeepSeek-V2, a strong Mixture-of-Experts (MoE) language model characterized by economical training and efficient inference. It comprises 236B total parameters, of which 21B are activated for each token, and supports a context…

Large Language Models (LLMs) have achieved remarkable success across many applications, with Mixture of Experts (MoE) models demonstrating great potential. Compared to traditional dense models, MoEs achieve better performance with less…

We study architectural and optimization techniques for sample-efficient language modeling under the constraints of the BabyLM 2025 shared task. Our model, BLaLM, replaces self-attention with a linear-time mLSTM token mixer and explores…

In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with…

LLMs have demonstrated remarkable proficiency in understanding tasks but continue to struggle with long-context comprehension, particularly with content located in the middle of extensive inputs. This limitation, known as the…

We study continued training and supervised fine-tuning (SFT) of a language model (LM) to make effective use of long-context information. We first establish a reliable evaluation protocol to guide model development -- instead of perplexity…

The maximum element of the vector output by the Softmax function approaches zero as the input vector size increases. Transformer-based language models rely on Softmax to compute attention scores, causing the attention distribution to…

We present MoE-MLA-RoPE, a novel architecture combination that combines Mixture of Experts (MoE) with Multi-head Latent Attention (MLA) and Rotary Position Embeddings (RoPE) for efficient language modeling. Our approach addresses the…

Attention mechanisms, which enable a neural network to accurately focus on all the relevant elements of the input, have become an essential component to improve the performance of deep neural networks. There are mainly two attention…

Large Language Models (LLMs) encounter significant performance bottlenecks in long-sequence tasks due to the computational complexity and memory overhead inherent in the self-attention mechanism. To address these challenges, we introduce…

We present the first comprehensive study of latent multi-head attention (MLA) for small language models, revealing interesting efficiency-quality trade-offs. Training 30M-parameter GPT models on 100,000 synthetic stories, we benchmark three…

Sparse attention methods exploit the inherent sparsity in attention to speed up the prefilling phase of long-context inference, mitigating the quadratic complexity of full attention computation. While existing sparse attention methods rely…

Machine translation systems achieve near human-level performance on some languages, yet their effectiveness strongly relies on the availability of large amounts of parallel sentences, which hinders their applicability to the majority of…