DailyPapers

Tweeting interesting papers submitted at your own at and link models/datasets/demos to it!

[DailyPapers @HuggingPapers]

Paper: https://t.co/pENRuCMjC6

Models, weights & game assets (Medieval Castle, Ancient Egypt, City Racing): https://t.co/hc3BiNHQWT

[DailyPapers @HuggingPapers]

Skywork AI's Matrix-Game 4.0

A playable game world rendered by a video model in real time: 480p at 60 FPS on a single RTX 5090, with a key-press-to-display latency of just 112ms.

[DailyPapers @HuggingPapers]

Paper: https://t.co/8d9sRntK5Q

HF paper page: https://t.co/8d9sRntK5Q

From Ant International researchers. Code and models would be a great addition to the Hugging Face Hub.

[DailyPapers @HuggingPapers]

PAR: Progress-Aligned Recurrence for length generalization in looped LMs

Loop k learns to predict the state after k operations, and recurrent states are rescaled to first-iteration norms. A 0.5B Qwen2.5 trained on 32 steps hits 95.5% at 512 steps—16x extrapolation.

[DailyPapers @HuggingPapers]

Alibaba Research and NTU released FocusGUI on Hugging Face

Paper: https://t.co/2TR4DuonZd

Project: https://t.co/M6anmH5LI5

[DailyPapers @HuggingPapers]

FocusGUI

A lightweight adapter that keeps omni-modal GUI agents focused on what matters at each step, avoiding trajectory drift without fine-tuning the backbone.

[DailyPapers @HuggingPapers]

Privacy-first: no camera images, just lane lines, bounding boxes, and traffic-light states.

Dive in: https://t.co/3jJEpAF258

[DailyPapers @HuggingPapers]

NVIDIA just released a driving simulation evaluation dataset on Hugging Face

91 multiview driving scenes with 11 cameras each, featuring World Scenario Map controls and text prompts.

[DailyPapers @HuggingPapers]

NVIDIA just released a real-time surgical world model on Hugging Face

It generates surgical video from action commands, enabling real-time simulation for training and evaluating surgical robots.

https://t.co/pFblyxQlj0

[DailyPapers @HuggingPapers]

Paper: https://t.co/Ie4QhxRuzB

EgoX-Data-6M and EgoX model are coming to Hugging Face: https://t.co/2bcjwoKPmU

[DailyPapers @HuggingPapers]

EgoXperience: World-Grounded Egocentric Representations from Human Experience

A full-stack paradigm for joint reconstruction of camera, scene, hands, and objects from egocentric video. Reduces camera, hand, and depth errors by 68.9%, 65.6%, and 42.9%, improves object pose by 28.0%.

[DailyPapers @HuggingPapers]

No model matched experts: top score 86.55% DAQ

Dataset on Hugging Face: https://t.co/MPVifUYxpN

Paper: https://t.co/vJQtQzunxl

[DailyPapers @HuggingPapers]

OR-License: can LLM agents do real operations research?

31 real client projects worth $3.68M
10 industries
An LLM-powered simulated client to question
Hidden checks against expert solutions clients already accepted

[DailyPapers @HuggingPapers]

Boosts accuracy from 62% to 72.8% on sentiment classification,
using only ~300 examples for training.

Check it out: https://t.co/pgQd6yXzDQ

[DailyPapers @HuggingPapers]

Microsoft just released Sentiment-Analysis-7B on Hugging Face

A fine-tuned Mistral 7B that classifies sentiment with no predefined labels,
using reinforcement learning and an LLM judge.

[DailyPapers @HuggingPapers]

Try the Diffusers pipeline: https://t.co/7NEqIysxoo

Original model card: https://t.co/TDJkYwViHd

[DailyPapers @HuggingPapers]

NVIDIA just released RFD4-Proteína on Hugging Face

A flow-matching model for biomolecular design.
Generates protein sequences and all-atom structures together, conditioned on your molecular context.

[DailyPapers @HuggingPapers]

Explore the full inventory & load metadata: https://t.co/a4qWxygob5

Includes 5 configs, annotated CIF records, and a reproducibility guide.

[DailyPapers @HuggingPapers]

NVIDIA just released RFD4-Proteína on Hugging Face

A massive 238GB collection of protein structures for biomolecular design training – from PDB and AlphaFold DB, including interfaces, antibodies, and TCR.

[DailyPapers @HuggingPapers]

Find it here: https://t.co/0gIQCZOauI

Also check the dataset: https://t.co/UnjvfOnS6u

[DailyPapers @HuggingPapers]

NVIDIA just released the H-Embodied surgical robotics model on Hugging Face

A single 16B model that predicts robot actions and endoscopic video for dVRK-style robots, serving as a world model, policy, and inverse dynamics model.