Grid-to-Token

Roman Bachmann, Jesse Allardice, David Mizrahi, Enrico Fini, Ouguzhan Fatih Kar, Elmira Amirloo, Alaaeldin El-Nouby, Amir Zamir, Afshin Dehghan

Image tokenization has enabled major advances in autoregressive image generation by providing compressed, discrete representations that are more efficient to process than raw pixels. While traditional approaches use 2D grid tokenization, recent methods like TiTok have shown that 1D tokenization can achieve high generation quality by eliminating grid redundancies. However, these methods typically use a fixed number of tokens and thus cannot adapt to an image's inherent complexity. We introduce FlexTok, a tokenizer that projects 2D images into variable-length, ordered 1D token sequences. For example, a 256x256 image can be resampled into anywhere from 1 to 256 discrete tokens, hierarchically and semantically compressing its information. By training a rectified flow model as the decoder and using nested dropout, FlexTok produces plausible reconstructions regardless of the chosen token sequence length. We evaluate our approach in an autoregressive generation setting using a simple GPT-style Transformer. On ImageNet, this approach achieves an FID<2 across 8 to 128 tokens, outperforming TiTok and matching state-of-the-art methods with far fewer tokens. We further extend the model to support to text-conditioned image generation and examine how FlexTok relates to traditional 2D tokenization. A key finding is that FlexTok enables next-token prediction to describe images in a coarse-to-fine"visual vocabulary", and that the number of tokens to generate depends on the complexity of the generation task.

Haichao Zhang, Yun Fu

Token-based video representation has emerged as a promising approach for enabling large language models (LLMs) to interpret video content. However, existing token reduction techniques, such as pruning and merging, often disrupt essential positional embeddings and rely on continuous visual tokens sampled from nearby pixels with similar spatial-temporal locations. By removing only a small fraction of tokens, these methods still produce relatively lengthy continuous sequences, which falls short of the extreme compression required to balance computational efficiency and token count in video LLMs. In this paper, we introduce the novel task of Extreme Short Token Reduction, which aims to represent entire videos using a minimal set of discrete tokens. We propose VQToken, a neural discrete token representation framework that (i) applies adaptive vector quantization to continuous ViT embeddings to learn a compact codebook and (ii) preserves spatial-temporal positions via a token hash function by assigning each grid-level token to its nearest codebook entry. On the Extreme Short Token Reduction task, our VQToken compresses sequences to just 0.07 percent of their original length while incurring only a 0.66 percent drop in accuracy on the NextQA-MC benchmark. It also achieves comparable performance on ActNet-QA, Long Video Bench, and VideoMME. We further introduce the Token Information Density (TokDense) metric and formalize fixed-length and adaptive-length subtasks, achieving state-of-the-art results in both settings. Our approach dramatically lowers theoretical complexity, increases information density, drastically reduces token counts, and enables efficient video LLMs in resource-constrained environments.

Stéphanie Rey, Anke M. Brock, Brygg Ullmer, Nadine Couture

Smart grid is envisioned to be the technology capable of scheduling user's energy requirement based on demand and decentralized nature. These challenges pose extreme pressure on finding advanced technologies and sustainable solutions for secure and reliable operations of the power system working inside the blockchain technology for managing exchange and trading of energy by means of specific tokens. For efficient utilization and functioning of the power grid we need a decentralised system which is transparent, trustless and makes transactions faster, there are a number of solutions proposed but none of them address the issue of transaction time in trade and penalty for defaulters. In this work we propose here an energy transaction network which implements blockchain technology for validating transaction of energy between producer/consumer or prosumer and saves energy and time using smart grids

J. Salzmann

Providing a global time reference for System-on-Chips (SoCs) is a challenging task. Conventional clock trees suffer from skews that unavoidably grow with the circuit size, which render them inappropriate for large circuits. In order to provide a synchronous abstraction at least locally, the globally asynchronous locally synchronous (GALS) approach can be used, which splits a circuit into several independently clocked islands and resorts to asynchronous communication between these islands. Several alternative approaches for achieving global synchrony were proposed in the past, some harnessing physical effects or other techniques, which suffer from different drawbacks. Inspired by certain self-timed asynchronous circuits, we present a fully digital distributed oscillator based on perpetual token exchange, which considerably outperforms alternative solutions in terms of the achievable local skew and simplicity. Furthermore, we experimentally study the fault-tolerance properties of our approach.