On Hugging Face

Archive, page 1

Wednesday 14 October 2026

Hugging Face Blog · 3h #

HiRO-ACEv2 simulates extreme weather at 3 km resolution cheaply

HiRO-ACEv2 is an AI weather model that simulates extreme weather at 3 km resolution for a fraction of the usual cost, released on Hugging Face.

HiROOpenOpen weights

Reddit · 4h #

Ling-3.1-flash: 560B-parameter MoE model with 1M context

An open Hugging Face model with 560B total parameters, 25B active per token and a 1M-token context window. It follows Ling 3.0, which was 127B.

inclusionAIOpenOpen weights

Tuesday 13 October 2026

Reddit · 1d #

Texture-fix VAE for Qwen Image 2.1 released on Hugging Face

A VAE that reduces checkerboard artifacts in Qwen Image 2.1 generations, shared on Hugging Face. Differences are only visible when zooming past 500%.

madebyollinOpenOpen weights

Reddit · 1d #

Aziib Pixel Style LoRA released for Krea2 and Anima

A pixel-art LoRA for image models Krea2 and Anima, downloadable from Civitai with the trigger word 'aziib pixel style'.

OpenOpen weights

HF Papers · 1d #

Urbex paper adds agentic spatial grounding for city-scale 3D

Urbex is a research framework that lets a vision-language agent actively search a city-scale 3D scene, querying landmarks and zooming instead of ranking precomputed instance proposals.

Paper · ▲ 50 upvotes

UrbexOpen

HF Papers · 1d #

Paper studies memory modality selection for GUI agents

A paper examines how GUI agents should choose the modality for retrieved memory, arguing existing agents use a fixed representation instead of adapting the format to the task.

Paper · ▲ 25 upvotes

Open

HF Papers · 1d #

VideoJev casts video understanding as decision tasks

VideoJev is a family of models that turn video understanding and interaction into structured decision tasks, adapting pretrained MLLMs with a shared prediction head instead of autoregressive text answers.

Paper · ▲ 23 upvotes

VideoJevOpen

HF Papers · 1d #

Jeva proposes System One models from first principles

Jeva is a paper arguing current system-one model designs miss four structural requirements, including distinguishing answer absence from uncertainty and permutation invariance, and proposes models built around them.

Paper · ▲ 39 upvotes

JevaOpen

HF Papers · 1d #

Agent-Reflex trains language agents to reflect and recover from failure

Agent-Reflex is a training framework that teaches language agents to reflect on noisy MCP tool feedback and user corrections, improving failure recovery in long-horizon tasks.

Paper · ▲ 41 upvotes

Open

HF Papers · 1d #

GreenWAM world foundation model targets robotics

GreenWAM is a world foundation model that both predicts how a robot's actions change a scene and chooses those actions, aiming to work across different robot bodies.

Paper · ▲ 28 upvotes

HF Papers · 1d #

Cepheid paper targets cross-family motion completion

A research paper on sparse-to-dense motion completion: recovering dense mesh motion from sparse vertex displacements. It shows residual priors trained on one geometry family often fail to beat parameter-free harmonic extension on other families.

Paper · ▲ 21 upvotes

HF Papers · 1d #

WAVE-0 paper bridges VLA policies and world action models

A research paper proposing lightweight latent future prediction to connect vision-language-action policies with world action models, aiming to model action consequences without costly video generation.

Paper · ▲ 20 upvotes

HF Papers · 1d #

Tev scales long-horizon agentic environments with dense rewards

Paper Tev presents a method to scale long-horizon agentic training environments using dense progress rewards, aiming to help open models work autonomously for hours to days.

Paper · ▲ 20 upvotes

Open

Monday 12 October 2026

HF Papers · 2d #

Multics paper proposes hierarchical communication for scalable multi-player world models

Multics is a research paper presenting a generative world model that simulates shared environments from multiple players' views. It uses hierarchical communication to avoid the quadratic cost of fully connected approaches.

Paper · ▲ 30 upvotes

HF Papers · 2d #

TPPO paper proposes tracking-value PPO to improve language model reasoning

TPPO is a research paper on reinforcement learning for language model reasoning. It tracks value evolution along trajectories to fix credit assignment when critics face sparse terminal rewards.

Paper · ▲ 23 upvotes

HF Papers · 2d #

X-Ragent paper adds tool-augmented RL for chest X-ray reasoning

Hugging Face paper X-Ragent presents tool-augmented reinforcement learning for evidence-grounded chest X-ray interpretation, teaching agents when to pull in auxiliary evidence.

Paper · ▲ 20 upvotes

HF Papers · 2d #

VEDJE compresses video features for scalable text retrieval

VEDJE is a joint encoder that compresses features within sampled video frames while keeping them separate in a reusable cache, speeding up video-text retrieval.

Paper · ▲ 20 upvotes

Open

HF Papers · 2d #

Paper proposes self-distillation with verbal regularization

The paper examines on-policy self-distillation where a teacher built from the same model uses privileged information, and proposes verbal regularization to curb bias from that supervision.

Paper · ▲ 20 upvotes

Saturday 10 October 2026