#Creative
Curated coverage on «Creative»: releases, research, benchmarks, open-source.
Apple developed TGPO to train video models to understand time
Apple ML Research introduced TGPO, a method that teaches video models to understand the order and evolution of events in egocentric video through reinforcement of…
Combinatorial Optimization in Robust Machine Learning: A Survey of Methods
A new survey on arXiv shows how combinatorial optimization methods provide formal guarantees of interpretability, robustness, and fairness for AI models — where…
Remote Sensing Foundation Models: How to Adapt AI for Satellite Data
Research shows that foundation models require specialized adaptation for satellite data, as measurement physics differs from standard computer vision.
Character.AI launches AI vertical video series c.ai Series
Meta opens access to Muse Spark 1.1 for AI coding through a new API
OpenAI releases the GPT-5.6 family with three models and an ultra mode
Infinity-Parser2 surpassed DeepSeek-OCR-2 in document parsing and opened a dataset of 5 million examples
TACO: a method removes credit “contamination” in RL training of language models
The new TACO method addresses a systemic problem in RL training for LLMs: unlikely “tail” tokens receive the same credit as correct ones, gradually degrading the…
New MiniLM-based method more accurately detects queries outside chatbots' knowledge scope
Researchers proposed multi-cluster boundary learning on MiniLM embeddings to identify out-of-scope queries in bots — the method outperformed all baseline solutions…
LiST: a new method simultaneously delivers accuracy, robustness, and calibration in neural networks
Hol-PCFG cut syntactic parsing parameters by 99.94% and took the lead in six languages
Datalab Lift vs Competitors: How a 9B Document Extractor Works with JSON Schema
Meta released Muse Spark 1.1 — multimodal agent with 1 million token context
GitLost vulnerability on GitHub: AI agents leak data from private repositories
Noma Labs researchers identified the GitLost vulnerability on GitHub — under certain prompts, AI agents extract data from private repositories and publish it in…
Google's SynthID watermark exposed the first high-profile deepfake of a U.S. senator
Snopes used Google's SynthID watermarking technology and confirmed that the viral photo of Senator McConnell in a hospital is an AI-generated deepfake.