Large Language Model
NetHack is Hard to Hack
In this work, we conduct a comprehensive study of NetHack and examine various learning mechanisms to enhance the performance of neural models.
DesCo: Learning Object Recognition with Rich Language Descriptions
Recent development in vision-language approaches has instigated a paradigm shift in learning visual recognition models from language supervision. These approaches align objects with language queries (e.g. "a photo of a cat") and thus improve the models' adaptability to novel objects and domains.
OpenAI has officially retired the controversial GPT-4o model
Samsung Galaxy Unpacked 2026 is Feb. 25 Valve's Steam Machine: Everything we know Some users are mourning GPT-4o's discontinuation on February 13, despite the concerns that the cult-favorite model was dangerously sycophantic. OpenAI's GPT-4o may have survived its first brush with going offline, but it won't be as lucky this time. OpenAI has officially retired GPT-4o, the ChatGPT model that was seen as more conversational and notoriously sycophantic, on February 13. The news of GPT-4o's end was first announced in a post on the OpenAI website in January, but the discontinuation also included GPT-5, GPT-4.1, However, a wave of user complaints led OpenAI to restore access to GPT-4o but with no guarantee that it'll be around forever.
Pandora's Box: Towards Building Universal Attackers against Real-World Large Vision-Language Models
Large Vision-Language Models (L VLMs) have demonstrated remarkable capabilities across a wide range of multimodal understanding tasks.
Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments
While humans accomplish 72.4% of the tasks, the best
U-DiTs: Downsample Tokens in U-Shaped Diffusion Transformers
DiTs (U-DiTs) in the paper and conduct extensive experiments to demonstrate the extraordinary performance of U-DiT models.
US military used Anthropic's AI model Claude in Venezuela raid, report says
A spokesperson for Anthropic declined to comment on whether Claude was used in the operation, but said any use of the tool was required to comply with its policies. A spokesperson for Anthropic declined to comment on whether Claude was used in the operation, but said any use of the tool was required to comply with its policies. US military used Anthropic's AI model Claude in Venezuela raid, report says Wall Street Journal says Claude used in operation via Anthropic's partnership with Palantir Technologies Sat 14 Feb 2026 11.15 ESTFirst published on Sat 14 Feb 2026 10.53 EST Claude, the AI model developed by Anthropic, was used by the US military during its operation to kidnap Nicolás Maduro from Venezuela, the Wall Street Journal revealed on Saturday, a high-profile example of how the US defence department is using artificial intelligence in its operations. The US raid on Venezuela involved bombing across the capital, Caracas, and the killing of 83 people, according to Venezuela's defence ministry. Anthropic's terms of use prohibit the use of Claude for violent ends, for the development of weapons or for conducting surveillance.
