The paper introduces K-Token Merging, a latent-space compression method that merges each contiguous block of K token embeddings into a single embedding using a lightweight encoder. The compressed sequence is then processed by a LoRA-adapted large language model, while generation continues in the original vocabulary. Experiments on tasks such as structural reasoning, sentiment classification, and code editing demonstrate that K-Token Merging achieves up to 75% input length reduction with minimal performance loss, placing it on the Pareto frontier of performance versus compression.
By Zihao Xu, John Harvill, Ziwei Fan, Yizhou Sun, Hao Ding, Hao Wang
arXiv:2607. 29378v1 Announce Type: cross Abstract: Large language models (LLMs) generate text by auto-regressively sampling the next token.
By Pirzada Suhail, Nagasai Saketh Naidu, Atanu R Sinha, Amit Sethi
Cola2 is a hierarchical language modeling framework that jointly optimizes continuous semantic compression, reconstruction–generation balance, and continuous–discrete co‑adaptation. It dynamically segments text into semantic spans, controls compression via negative feedback, and trains a block‑causal masked diffusion decoder with clean, noisy, and dropped latent conditions. Experiments on five downstream tasks with 210 billion training tokens show competitive scaling, and an unconverged checkpoint with 346 billion source tokens outperforms Llama‑3‑8B on HellaSwag and StoryCloze.
By Hongcan Guo, Yian Zhao, Feng Wang, Shen Nie, Guoqiang Wei, Yan Zeng, Hengshuang Zhao
arXiv:2610.33811v1 Announce Type: new
Abstract: Continuous-state flow and diffusion models offer a promising approach to fast, parallel language generation, but matching the generation quality of aut...
By Baiyu Su, Runlong Liao, Lizhang Chen, Qiang Liu
The paper introduces LLMAE, a technique that transforms a pretrained decoder-only language model into a continuous text autoencoder by inserting a fixed-length latent bottleneck into its internal activations. Using a 270M Gemma 3 model with structured attention masks, LoRA adaptation, and KL regularization, LLMAE achieves near-perfect reconstruction of text sequences up to 1024 tokens. The authors further show that the resulting latent representation can be leveraged to train a latent text diffusion model for detailed image captioning, demonstrating downstream utility.
By Arkanath Pathak, Unnat Jain, Alexander C. Berg
SCOPE is a training‑free generative prompt‑compression framework that reduces LLM input length by chunking a prompt into semantically coherent segments, rewriting each chunk to be more concise, and then reconstructing a coherent prompt. Unlike token‑removal methods, SCOPE’s chunk‑level rewriting preserves critical information and text coherence, and includes optimization techniques for finer‑grained control of compression ratios. Extensive evaluations on question‑answering and summarization tasks show that SCOPE consistently outperforms selective compression baselines, especially at high compression ratios.
By Tinghui Zhang, Yifan Wang, Daisy Zhe Wang
The paper investigates why token prediction, a common pre‑training objective for language models, yields useful representations. It introduces a statistical framework linking token prediction accuracy to the geometry of token embeddings, showing that accurate predictions organize embeddings according to Hellinger distances between context distributions. The authors also propose a self‑consistency principle that refines contextual representations through repeated application of a shared block, and provide downstream guarantees for token generation, community recovery, and linear classification.
By Shulei Wang
arXiv:2609.15338v1 Announce Type: cross
Abstract: Large Language Models (LLMs) primarily perform inference at the token level, resulting in substantial memory overhead and compromised computational e...
By Peipei Li, Dongsen Zhang, Yuchen Liu, Wenjun Xu
Sentence-level AI-generated text detection (S-AGTD) for hybrid documents, where humans and LLMs co-author one text, faces two gaps: existing methods classify each sentence in isolation, discarding inter-sentence dependencies, and existing benchmarks omit the newest generation of generators. We construct MOSAIC, a benchmark of 16,000 hybrid documents over PubMed and XSum, generated by DeepSeek-V3.
The paper introduces Seq-Drifting, a new framework that enables diffusion language models to generate an entire text sequence in a single forward pass. By applying a drift field to continuous token representations, the model learns a direct mapping from Gaussian noise to text, using attraction to guide outputs toward contextually plausible tokens and repulsion to prevent repetition and collapse. Experiments show that Seq-Drifting achieves a perplexity of 36.01 on LM1B and performs competitively in unconditional, conditional, translation, summarization, and reasoning tasks.
By Jiayu Liu, Ruizhi Zhao, Tianyun Ji, Yee Whye Teh, Zhenya Huang, Ning Miao
The paper introduces SCSP, a training‑free framework that improves long‑context embeddings by selectively pooling informative tokens. SCSP partitions documents into sentence‑aware chunks, adds a semantic compression prompt to each chunk, and uses prompt‑isolated attention masks to estimate token importance. The selected tokens’ intermediate‑layer representations are aggregated to form the final embedding, yielding consistent performance gains across zero‑shot and fine‑tuned models on long‑context benchmarks.
By Zifeng Cheng, Jie Zheng, Zhiwei Jiang, Shuwen Wang, Fei Shen, Shiping Ge, Qing Gu
The paper introduces Representation-based Masked Diffusion Model (RMDM), a new framework for language modeling that improves upon existing Masked Diffusion Models by incorporating global semantic guidance. RMDM encodes text into a continuous semantic space with a pretrained encoder, normalizes this representation to a Gaussian prior via an invertible transformation, and then trains a masked diffusion model conditioned on this latent representation to coordinate parallel token updates. Experiments show that RMDM yields higher generation quality, especially when using aggressive few‑step sampling.
By Yangrong Hu, Ding Huang, Xueyu Zhou, Jian Huang