Tag
BranchShine-CR is a compact 25M-parameter model for multilingual IPA transcription that uses log-mel features, an E-Branchformer encoder, self-conditioned CTC, and consistency regularization, achieving a 22.3% relative reduction in IPA character error rate compared to prior models.
The paper introduces TAPe+MLv3, a compact computer vision system using structured representation for multi-task tasks, achieving competitive performance on benchmarks like COCO with fewer than 100,000 parameters.
Introduces Spark-X2.5-1.7B, a compact text generation model designed for smooth bilingual conversations in English and Chinese, with early positive reception.
DreamX-Creator 1.0 is a compact 7B native joint audio-video generator that uses cross-modal attention, progressive training, and reinforcement learning to produce synchronized 2K resolution outputs, aiming to democratize high-quality audio-video generation.
Official announcement of Horus Cyber Nano 1.0, showcasing benchmark results, architecture details, and confirming the release date for model weights.
ALPHABET is a compact linear-time sequence model that compresses temporal history into stable complex pole modes, achieving competitive performance with fewer parameters and faster inference.
DistilVDR is a compact 524M visual document retriever distilled from an 8B teacher via cosine alignment, achieving near-teacher accuracy on ViDoRe with 15.6x smaller indexes and faster indexing.
Nanbeige4.2-3B is a compact 3B parameter general agentic model pretrained from scratch with a Looped Transformer, achieving strong agentic and reasoning performance, outperforming larger models on diverse benchmarks. The model and code are open-sourced.
Nanbeige4.2-3B is a compact agentic model using Looped Transformer architecture, demonstrating strong performance on agentic and reasoning benchmarks at the 3B scale, outperforming larger models such as Qwen3.5-9B and Gemma4-12B.
FreyaTTS is a compact, tokenizer-free Turkish-first text-to-speech model based on a non-autoregressive conditional flow-matching Diffusion Transformer, achieving state-of-the-art performance with a fraction of the parameters of larger systems and released under Apache-2.0.
GRAPE is a training framework that progressively exposes parameter space during adversarial training, achieving higher robust accuracy with fewer parameters compared to fixed-structure methods on CIFAR-10.
VTXAI has released Helvete nano, a compact 2B parameter language model designed for unrestricted conversations and creative freedom.
This paper presents CosmicFish-HRM, a compact 82.77M parameter language model with a hierarchical reasoning module that dynamically allocates reasoning compute during inference, learning when to halt based on input complexity.
Lens is a compact 3.8B-parameter text-to-image model from Microsoft that achieves competitive performance with larger models while requiring significantly less training compute, using dense captions, multi-resolution batching, and efficient architecture.
Google introduces Gemma 3 270M, a compact 270-million parameter model designed for efficient on-device AI with strong instruction-following capabilities and extreme energy efficiency (0.75% battery for 25 conversations on Pixel 9 Pro).
Arch-Router is a compact 1.5B model that aligns LLM routing with human preferences by mapping queries to user-defined domains and action types, outperforming proprietary models in subjective evaluations.